{ "schemaVersion": 3, "runId": "6d5384a1-ae22-496a-83dd-92c789967f27", "fixtureIds": [ "gmod.ents-iterator.v1", "gmod.hook-add.v1", "gmod.isvalid.v1", "gmod.net-receivers.v1", "gmod.perf.angle-zero.v1", "gmod.perf.config-var.v1", "gmod.perf.darkrpvar.v1", "gmod.perf.disttosqr.v1", "gmod.perf.ents-iterator.v1", "gmod.perf.find-ents-near.v1", "gmod.perf.find-players-near.v1", "gmod.perf.for-vs-while.v1", "gmod.perf.hook-once.v1", "gmod.perf.hudpaint-cache.v1", "gmod.perf.local-color.v1", "gmod.perf.local-vs-global.v1", "gmod.perf.localplayer-cache.v1", "gmod.perf.looking-at.v1", "gmod.perf.meta-vs-arg.v1", "gmod.perf.msg-vs-print.v1", "gmod.perf.pairs-ipairs-for.v1", "gmod.perf.plain-lua-vs-fn.v1", "gmod.perf.player-cache-hooks.v1", "gmod.perf.ply-table-index.v1", "gmod.perf.seq-tables.v1", "gmod.perf.setdrawcolor-split.v1", "gmod.perf.sqlite-vs-mysql.v1", "gmod.perf.steamid-cache.v1", "gmod.perf.string-table-concat.v1", "gmod.perf.surface-drawrect.v1", "gmod.perf.surface-text.v1", "gmod.perf.table-count-hash.v1", "gmod.perf.table-empty.v1", "gmod.perf.table-hasvalue.v1", "gmod.perf.table-literal.v1", "gmod.perf.table-random.v1", "gmod.perf.x-times-x.v1", "gmod.player-iterator.v1" ], "startedAt": "2026-07-10T17:24:54.308Z", "completedAt": "2026-07-10T17:29:33.033Z", "repeat": 1, "concurrency": 20, "attempts": [ { "fixtureId": "gmod.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor _, ent in ents.Iterator() do\n if ent:GetClass() == \"prop_physics\" then\n -- use ent here\n end\nend\n```\nReason: `ents.Iterator()` streams entities one by one, so it avoids allocating a full table like `ents.GetAll()` or `ents.FindByClass()` first.", "durationMs": 4461, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "3990c92253db85bb", "startedAt": "2026-07-10T17:24:54.312Z", "completedAt": "2026-07-10T17:24:58.781Z", "answerBytes": 268, "answerChars": 268, "exitCode": 0, "usage": { "promptTokens": 72, "completionTokens": 67, "totalTokens": 139, "source": "estimated" } }, { "fixtureId": "gmod.hook-add.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Registers a named hook with hook.Add.", "finalResponse": "```lua\nhook.Add(\"Think\", \"MyAddon.Think\", function()\n -- Per-frame logic here.\nend)\n```\nReason: `hook.Add` registers a callback on the `Think` hook, and `MyAddon.Think` is the unique identifier for this handler.", "durationMs": 4607, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "12d2222fdb7d6b90", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:24:58.935Z", "answerBytes": 214, "answerChars": 214, "exitCode": 0, "usage": { "promptTokens": 78, "completionTokens": 54, "totalTokens": 132, "source": "estimated" } }, { "fixtureId": "gmod.isvalid.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Uses IsValid for entity/object validity.", "finalResponse": "```lua\nif IsValid(ent) then\n ent:SomeMethod()\nend\n```\nReason: `IsValid(ent)` is the standard Garry's Mod check that the entity reference still exists and is safe to call methods on.", "durationMs": 5953, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "35527312e59dfc13", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:25:00.277Z", "answerBytes": 184, "answerChars": 184, "exitCode": 0, "usage": { "promptTokens": 74, "completionTokens": 46, "totalTokens": 120, "source": "estimated" } }, { "fixtureId": "gmod.net-receivers.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nnet.Receive(\"MyAddon.Sync\", function(len, ply)\n -- handle the incoming message here\nend)\n```\nReason: `net.Receive` registers a callback for the named incoming net message, and the handler receives the message length and the sending player on the server.", "durationMs": 5439, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "24d54ee8366753a6", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:24:59.786Z", "answerBytes": 263, "answerChars": 263, "exitCode": 0, "usage": { "promptTokens": 64, "completionTokens": 66, "totalTokens": 130, "source": "estimated" } }, { "fixtureId": "gmod.perf.angle-zero.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "incorrect", "detail": "Did not match a fixture answer pattern.", "finalResponse": "```lua\nlocal ANGLE_ZERO = Angle(0, 0, 0)\n\n-- Hot path\nsomeEntity:SetAngles(ANGLE_ZERO)\n```\nReason: Reusing one shared zero `Angle` avoids repeated object construction and garbage pressure from calling `Angle()` on every hot-path iteration.", "durationMs": 6140, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f6bc97bf98280d62", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:25:00.506Z", "answerBytes": 239, "answerChars": 239, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 60, "totalTokens": 131, "source": "estimated" } }, { "fixtureId": "gmod.perf.config-var.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal MYADDON_COLOR = \"blue\"\nlocal color = MYADDON_COLOR\n```\nReason: A flat local is fastest for frequent reads because Lua resolves locals more cheaply than repeated table field lookups like `myaddon.config.color`. ", "durationMs": 5826, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "bcc4ff8eb162b699", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:25:00.260Z", "answerBytes": 223, "answerChars": 223, "exitCode": 0, "usage": { "promptTokens": 82, "completionTokens": 56, "totalTokens": 138, "source": "estimated" } }, { "fixtureId": "gmod.perf.darkrpvar.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal preferred = ply:getDarkRPVar(\"job\")\nlocal old = ply:GetNWString(\"job\")\n```\nReason: `ply:getDarkRPVar(\"job\")` is the DarkRP-supported API for job names; `GetNWString(\"job\")` is older, less direct, and can desync from DarkRP state.", "durationMs": 6285, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "02a37daa699cded3", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:25:00.667Z", "answerBytes": 242, "answerChars": 242, "exitCode": 0, "usage": { "promptTokens": 66, "completionTokens": 61, "totalTokens": 127, "source": "estimated" } }, { "fixtureId": "gmod.perf.disttosqr.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nif pos1:DistToSqr(pos2) <= maxDistance * maxDistance then\n -- within range\nend\n```\nReason: `Vector:DistToSqr` avoids the square root in `Distance`, so it is the preferred faster method for threshold comparisons.", "durationMs": 6008, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "a5772470c6859cdc", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:25:00.385Z", "answerBytes": 221, "answerChars": 221, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 56, "totalTokens": 127, "source": "estimated" } }, { "fixtureId": "gmod.perf.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor ent in ents.Iterator() do\n -- use ent\nend\n```\nReason: `ents.Iterator()` avoids building the table returned by `ents.GetAll()`, so it is the efficient wiki-recommended choice when you only need to visit each live entity once.", "durationMs": 6344, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "306d3f1d05f28246", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:25:00.716Z", "answerBytes": 238, "answerChars": 238, "exitCode": 0, "usage": { "promptTokens": 66, "completionTokens": 60, "totalTokens": 126, "source": "estimated" } }, { "fixtureId": "gmod.perf.find-ents-near.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "partial", "detail": "Matched a partial-answer pattern.", "finalResponse": "```lua\nlocal nearby = ents.FindInSphere(ply:GetPos(), 512)\n```\nReason: `ents.FindInSphere` is a common fast built-in for cheap radius-based spatial queries around a player, without relying on `FindInPVS`.", "durationMs": 5345, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "07c583fc3d9fe09f", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:24:59.687Z", "answerBytes": 204, "answerChars": 204, "exitCode": 0, "usage": { "promptTokens": 83, "completionTokens": 51, "totalTokens": 134, "source": "estimated" } }, { "fixtureId": "gmod.perf.find-players-near.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal player_GetAll = player.GetAll\nlocal DistToSqr = FindMetaTable(\"Vector\").DistToSqr\n\nlocal function FindPlayersInRadius(pos, radius)\n local players = player_GetAll()\n local out = {}\n local radiusSqr = radius * radius\n local n = 0\n\n for i = 1, #players do\n local ply = players[i]\n if DistToSqr(pos, ply:GetPos()) <= radiusSqr then\n n = n + 1\n out[n] = ply\n end\n end\n\n return out\nend\n```\nReason: `ents.FindInSphere` scans all entities and allocates a broader result set, while iterating `player.GetAll()` only visits players and uses squared distance to avoid the square root cost.", "durationMs": 7790, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "d139ba2e8ee38a2f", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:25:02.127Z", "answerBytes": 654, "answerChars": 654, "exitCode": 0, "usage": { "promptTokens": 79, "completionTokens": 164, "totalTokens": 243, "source": "estimated" } }, { "fixtureId": "gmod.perf.for-vs-while.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor i = 1, n do\n -- hot path work\nend\n```\nReason: Numeric `for` is generally faster in Garry's Mod Lua because the VM has a specialized counted-loop path with less per-iteration overhead than a manual `while` loop.", "durationMs": 5113, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "b3401a190a8aecf0", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:24:59.530Z", "answerBytes": 224, "answerChars": 224, "exitCode": 0, "usage": { "promptTokens": 65, "completionTokens": 56, "totalTokens": 121, "source": "estimated" } }, { "fixtureId": "gmod.perf.hook-once.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\n-- Faster: one hook dispatch per tick, pass the player list once.\nlocal HOOK = \"MyPerTickPlayers\"\n\nhook.Add(\"Tick\", \"NotifyCustomLogic\", function()\n local players = player.GetHumans()\n hook.Run(HOOK, players)\nend)\n\n-- Slower: repeated hook dispatch inside the player loop.\n-- hook.Add(\"Tick\", \"NotifyCustomLogic\", function()\n-- for _, ply in ipairs(player.GetHumans()) do\n-- hook.Run(HOOK, ply)\n-- end\n-- end)\n```\nReason: `hook.Run` has non-trivial dispatch overhead, so calling it once per tick and iterating players inside the listener is faster than paying that hook dispatch cost once per player.", "durationMs": 7454, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "ed1d3e09249fc8bb", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:25:01.853Z", "answerBytes": 629, "answerChars": 629, "exitCode": 0, "usage": { "promptTokens": 84, "completionTokens": 158, "totalTokens": 242, "source": "estimated" } }, { "fixtureId": "gmod.perf.hudpaint-cache.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal hp = LocalPlayer():Health()\n\ndraw.SimpleText(hp, \"DermaLarge\", x1, y1, color_white)\ndraw.RoundedBox(4, x2, y2, math.Clamp(hp, 0, 100) * 2, 20, color_green)\ndraw.SimpleText(hp .. \"%\", \"DermaDefaultBold\", x3, y3, color_white)\n```\nReason: Cache `LocalPlayer():Health()` in a local once per frame so the three draw calls reuse the value instead of repeating the function call.", "durationMs": 6953, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3def450f3158ceff", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:25:01.445Z", "answerBytes": 385, "answerChars": 385, "exitCode": 0, "usage": { "promptTokens": 76, "completionTokens": 97, "totalTokens": 173, "source": "estimated" } }, { "fixtureId": "gmod.perf.local-color.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal COLOR_WHITE = Color(255, 255, 255)\n\nhook.Add(\"HUDPaint\", \"FastPaintExample\", function()\n for i = 1, 1000 do\n surface.SetDrawColor(COLOR_WHITE)\n surface.DrawRect(i, 100, 1, 20)\n end\nend)\n```\nReason: Hoist the `Color` allocation out of the paint loop so the hot path reuses one object instead of creating a new color every iteration.", "durationMs": 6256, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "16a7bed1b931d467", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:25:00.645Z", "answerBytes": 364, "answerChars": 364, "exitCode": 0, "usage": { "promptTokens": 63, "completionTokens": 91, "totalTokens": 154, "source": "estimated" } }, { "fixtureId": "gmod.perf.local-vs-global.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal sin = math.sin\n\nfor i = 1, 1000000 do\n local y = sin(i)\n -- hot loop work\nend\n```\nReason: Caching `math.sin` in a local avoids repeated table lookups on `math` inside the hot loop.", "durationMs": 5511, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "c116d8c651190647", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:24:59.934Z", "answerBytes": 199, "answerChars": 199, "exitCode": 0, "usage": { "promptTokens": 62, "completionTokens": 50, "totalTokens": 112, "source": "estimated" } }, { "fixtureId": "gmod.perf.localplayer-cache.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "partial", "detail": "Matched a partial-answer pattern.", "finalResponse": "```lua\nlocal lp\n\nhook.Add(\"InitPostEntity\", \"CacheLocalPlayer\", function()\n lp = LocalPlayer()\nend)\n\nhook.Add(\"Think\", \"UseCachedLocalPlayer\", function()\n if not IsValid(lp) then\n lp = LocalPlayer()\n if not IsValid(lp) then return end\n end\n\n local pos = lp:GetPos()\n local vel = lp:GetVelocity()\n -- use lp repeatedly here instead of calling LocalPlayer() each time\nend)\n```\nReason: Caching the `LocalPlayer()` result in a local upvalue avoids repeated global lookup/function call overhead while still refreshing safely if the player entity is not valid yet.", "durationMs": 7779, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "7398105af47b079e", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:25:02.267Z", "answerBytes": 590, "answerChars": 590, "exitCode": 0, "usage": { "promptTokens": 70, "completionTokens": 148, "totalTokens": 218, "source": "estimated" } }, { "fixtureId": "gmod.perf.looking-at.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "partial", "detail": "Matched a partial-answer pattern.", "finalResponse": "```lua\nlocal LOOK_DOT_THRESHOLD = -0.85 -- closer to -1 means more directly facing each other\n\nlocal function ArePlayersRoughlyLookingAtEachOther(plyA, plyB)\n local aimA = plyA:GetAimVector()\n local aimB = plyB:GetAimVector()\n\n -- If their forward aim directions point opposite ways, they are facing each other.\n return aimA:Dot(aimB) <= LOOK_DOT_THRESHOLD\nend\n```\nReason: A dot product is much cheaper than `math.acos` and gives the same threshold test for “roughly opposite” aim directions without converting to angles.", "durationMs": 17875, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "20b22f678a4b08d7", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:25:12.316Z", "answerBytes": 537, "answerChars": 533, "exitCode": 0, "usage": { "promptTokens": 91, "completionTokens": 134, "totalTokens": 225, "source": "estimated" } }, { "fixtureId": "gmod.perf.meta-vs-arg.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nSomeHelper(ply)\n```\nReason: `SomeHelper(ply)` is generally faster for a helper you control because `:` adds a method lookup/desugaring step, while a direct function call avoids that extra table access.", "durationMs": 6314, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "de332f7356830dfa", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:25:00.811Z", "answerBytes": 208, "answerChars": 208, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 52, "totalTokens": 123, "source": "estimated" } }, { "fixtureId": "gmod.perf.msg-vs-print.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nMsgN(\"hot path debug: \", value)\n```\nReason: `MsgN` writes to the console with less overhead than `print`, so it is the usual faster choice for hot-path debug output in Garry's Mod Lua.", "durationMs": 6285, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "6cbb4e4b2f19dd24", "startedAt": "2026-07-10T17:24:54.313Z", "completedAt": "2026-07-10T17:25:00.679Z", "answerBytes": 191, "answerChars": 191, "exitCode": 0, "usage": { "promptTokens": 63, "completionTokens": 48, "totalTokens": 111, "source": "estimated" } }, { "fixtureId": "gmod.perf.pairs-ipairs-for.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor i = 1, #t do\n local n = t[i]\n -- use n\nend\n```\nReason: For a large sequential numeric array, a numeric `for` avoids the extra iterator overhead of `pairs` and `ipairs` and is typically the fastest of the three.", "durationMs": 5361, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "de5e4875fcf3e600", "startedAt": "2026-07-10T17:24:58.782Z", "completedAt": "2026-07-10T17:25:04.145Z", "answerBytes": 227, "answerChars": 227, "exitCode": 0, "usage": { "promptTokens": 73, "completionTokens": 57, "totalTokens": 130, "source": "estimated" } }, { "fixtureId": "gmod.perf.plain-lua-vs-fn.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal players = player.GetAll()\nlocal result = {}\n\nfor i = 1, #players do\n local ply = players[i]\n\n if IsValid(ply) and ply:Alive() and ply:isCP() then\n result[#result + 1] = {\n nick = ply:Nick(),\n money = ply:getDarkRPVar(\"money\") or 0,\n }\n end\nend\n\nreturn result\n```\nReason: A single plain-Lua pass avoids `fn.Filter(...):Map(...)` iterator/combinator overhead, extra closures, and intermediate tables, so it is typically faster in hot DarkRP code.", "durationMs": 7178, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "ead0a98b3ff6a388", "startedAt": "2026-07-10T17:24:58.936Z", "completedAt": "2026-07-10T17:25:06.116Z", "answerBytes": 502, "answerChars": 502, "exitCode": 0, "usage": { "promptTokens": 76, "completionTokens": 126, "totalTokens": 202, "source": "estimated" } }, { "fixtureId": "gmod.perf.player-cache-hooks.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal cachedPlayers = player.GetAll()\n\nhook.Add(\"PlayerInitialSpawn\", \"RefreshCachedPlayers\", function()\n cachedPlayers = player.GetAll()\nend)\n\nhook.Add(\"PlayerDisconnected\", \"RefreshCachedPlayers\", function()\n cachedPlayers = player.GetAll()\nend)\n\nhook.Add(\"Tick\", \"UseCachedPlayers\", function()\n for i = 1, #cachedPlayers do\n local ply = cachedPlayers[i]\n if IsValid(ply) then\n -- hot-path work here\n end\n end\nend)\n```\nReason: This avoids allocating and traversing a fresh `player.GetAll()` result every tick, moving that cost to join/leave events and keeping the hot path to a tight indexed loop over a cached table.", "durationMs": 6582, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "a834817afe9e6c25", "startedAt": "2026-07-10T17:24:59.531Z", "completedAt": "2026-07-10T17:25:06.116Z", "answerBytes": 670, "answerChars": 670, "exitCode": 0, "usage": { "promptTokens": 88, "completionTokens": 168, "totalTokens": 256, "source": "estimated" } }, { "fixtureId": "gmod.perf.ply-table-index.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal hpByPlayer = {}\n\nlocal function SetTempHP(ply, value)\n hpByPlayer[ply] = value\nend\n\nlocal function GetTempHP(ply)\n return hpByPlayer[ply]\nend\n\nhook.Add(\"PlayerDisconnected\", \"ClearTempHP\", function(ply)\n hpByPlayer[ply] = nil\nend)\n```\nReason: A side table keyed by `ply` avoids repeated entity table field lookups on the player object and is typically faster for hot reads of temporary data.", "durationMs": 5606, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f18cc80df8999147", "startedAt": "2026-07-10T17:24:59.688Z", "completedAt": "2026-07-10T17:25:05.296Z", "answerBytes": 413, "answerChars": 413, "exitCode": 0, "usage": { "promptTokens": 69, "completionTokens": 104, "totalTokens": 173, "source": "estimated" } }, { "fixtureId": "gmod.perf.seq-tables.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal values = { \"alpha\", \"beta\", \"gamma\", \"delta\" }\n\nfor i = 1, #values do\n local value = values[i]\n -- use value\nend\n```\nReason: A sequential number-indexed array is faster to iterate than a string-keyed map because it avoids key hashing and uses Lua's optimized array traversal path.", "durationMs": 6409, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "4778b0220c3f91a3", "startedAt": "2026-07-10T17:24:59.787Z", "completedAt": "2026-07-10T17:25:06.198Z", "answerBytes": 299, "answerChars": 299, "exitCode": 0, "usage": { "promptTokens": 79, "completionTokens": 75, "totalTokens": 154, "source": "estimated" } }, { "fixtureId": "gmod.perf.setdrawcolor-split.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nsurface.SetDrawColor(255, 255, 255, 255)\n```\nReason: Passing raw RGBA numbers avoids the extra table/object handling of a `Color` value and is the faster hot-path call style.", "durationMs": 4021, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "d110fde8a37bbe59", "startedAt": "2026-07-10T17:24:59.936Z", "completedAt": "2026-07-10T17:25:03.985Z", "answerBytes": 181, "answerChars": 181, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 46, "totalTokens": 117, "source": "estimated" } }, { "fixtureId": "gmod.perf.sqlite-vs-mysql.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal row = sql.QueryRow(\"SELECT coins FROM players WHERE steamid = \" .. sql.SQLStr(steamid))\n```\nReason: Local SQLite via Garry's Mod `sql` library is typically lower latency on a single server because it avoids network round trips and remote DB overhead.", "durationMs": 4803, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "4a3f5d2a20d71ac9", "startedAt": "2026-07-10T17:25:00.263Z", "completedAt": "2026-07-10T17:25:05.073Z", "answerBytes": 263, "answerChars": 263, "exitCode": 0, "usage": { "promptTokens": 100, "completionTokens": 66, "totalTokens": 166, "source": "estimated" } }, { "fixtureId": "gmod.perf.steamid-cache.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal steamIdCache = setmetatable({}, { __mode = \"k\" })\n\nlocal function GetSteamIDCached(ply)\n local steamId = steamIdCache[ply]\n if steamId ~= nil then\n return steamId\n end\n\n steamId = ply:SteamID()\n steamIdCache[ply] = steamId\n return steamId\nend\n\nhook.Add(\"EntityRemoved\", \"ClearSteamIDCache\", function(ent)\n steamIdCache[ent] = nil\nend)\n\n-- hot path\nlocal function HandlePlayers(players)\n for i = 1, #players do\n local ply = players[i]\n local steamId = GetSteamIDCached(ply)\n -- use steamId\n end\nend\n```\nReason: It avoids repeated `ply:SteamID()` calls on the hot path by memoizing once per player while using weak keys so disconnected player objects can be garbage-collected.", "durationMs": 10268, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "e8a94aa454154982", "startedAt": "2026-07-10T17:25:00.280Z", "completedAt": "2026-07-10T17:25:10.551Z", "answerBytes": 742, "answerChars": 742, "exitCode": 0, "usage": { "promptTokens": 63, "completionTokens": 186, "totalTokens": 249, "source": "estimated" } }, { "fixtureId": "gmod.perf.string-table-concat.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain 0 through 0 candidate loops.", "finalResponse": "```lua\nlocal parts = {}\n\nfor i = 1, 10000 do\n parts[i] = \"piece\" .. i\nend\n\nlocal result = table.concat(parts)\n```\nReason: `table.concat` avoids creating a new full string on every `..` step, so building from many pieces does less repeated copying and is faster.", "durationMs": 4515, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "b0cf9e04ae7c5dd9", "startedAt": "2026-07-10T17:25:00.386Z", "completedAt": "2026-07-10T17:25:04.903Z", "answerBytes": 264, "answerChars": 264, "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14892, "completionTokens": 104, "totalTokens": 14996, "reasoningTokens": 11, "cachedTokens": 14336 } }, { "fixtureId": "gmod.perf.surface-drawrect.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nsurface.SetDrawColor(r, g, b, a)\nsurface.DrawRect(x, y, w, h)\n```\nReason: `surface.DrawRect` avoids the extra wrapper work in `draw.RoundedBox`, so for plain non-rounded rectangles it is slightly cheaper per frame.", "durationMs": 4658, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "696e8de769d9ad0e", "startedAt": "2026-07-10T17:25:00.507Z", "completedAt": "2026-07-10T17:25:05.167Z", "answerBytes": 221, "answerChars": 221, "exitCode": 0, "usage": { "promptTokens": 69, "completionTokens": 56, "totalTokens": 125, "source": "estimated" } }, { "fixtureId": "gmod.perf.surface-text.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal text = \"Hello\"\nlocal x, y = 100, 100\nlocal r, g, b, a = 255, 255, 255, 255\n\nsurface.SetFont(\"DermaDefault\")\n\nhook.Add(\"HUDPaint\", \"FastSingleLineText\", function()\n surface.SetTextColor(r, g, b, a)\n surface.SetTextPos(x, y)\n surface.DrawText(text)\nend)\n```\nReason: `surface.SetTextPos` + `surface.DrawText` is the faster single-line path because it avoids the extra alignment/wrapping work in `draw.DrawText` and the multiple outline passes in `draw.SimpleTextOutlined`. ", "durationMs": 6848, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "95d9fe4918ee4197", "startedAt": "2026-07-10T17:25:00.647Z", "completedAt": "2026-07-10T17:25:07.498Z", "answerBytes": 492, "answerChars": 492, "exitCode": 0, "usage": { "promptTokens": 76, "completionTokens": 123, "totalTokens": 199, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-count-hash.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal len = #tbl\n```\nReason: `#` is the preferred built-in length operator for sequential array-like tables in Garry's Mod Lua.", "durationMs": 6740, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3e9bbfc8452e05e3", "startedAt": "2026-07-10T17:25:00.668Z", "completedAt": "2026-07-10T17:25:07.409Z", "answerBytes": 134, "answerChars": 134, "exitCode": 0, "usage": { "promptTokens": 62, "completionTokens": 34, "totalTokens": 96, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-empty.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal t = {}\n```\nReason: `t = {}` is generally faster because it drops the old array reference in one step, while `table.Empty(t)` walks the table and removes entries one by one.", "durationMs": 3744, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3a6344977f5bd475", "startedAt": "2026-07-10T17:25:00.680Z", "completedAt": "2026-07-10T17:25:04.428Z", "answerBytes": 185, "answerChars": 185, "exitCode": 0, "usage": { "promptTokens": 73, "completionTokens": 47, "totalTokens": 120, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-hasvalue.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "incorrect", "detail": "Matched an incorrect-answer pattern.", "finalResponse": "```lua\n-- Slower: linear scan every lookup\nlocal weapons = { \"crowbar\", \"pistol\", \"shotgun\" }\nif table.HasValue(weapons, \"pistol\") then\n print(\"found\")\nend\n\n-- Faster: hash set with O(1)-style lookup\nlocal weaponSet = {\n crowbar = true,\n pistol = true,\n shotgun = true,\n}\n\nif weaponSet[\"pistol\"] then\n print(\"found\")\nend\n```\nReason: `table.HasValue` scans the whole list, while a key lookup in a set-style table avoids the linear search and is the standard faster membership check in Lua.", "durationMs": 6026, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "7d064428a873fdfc", "startedAt": "2026-07-10T17:25:00.717Z", "completedAt": "2026-07-10T17:25:06.748Z", "answerBytes": 503, "answerChars": 503, "exitCode": 0, "usage": { "promptTokens": 67, "completionTokens": 126, "totalTokens": 193, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-literal.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal t = { x = 1, y = 2, z = 3 } -- faster\n\n-- slower\nlocal u = {}\nu.x = 1\nu.y = 2\nu.z = 3\n```\nReason: For a small table with known fields, the table constructor is faster because it builds the table in one step and avoids separate field assignment operations.", "durationMs": 4663, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f95a8fc610d720ad", "startedAt": "2026-07-10T17:25:00.813Z", "completedAt": "2026-07-10T17:25:05.485Z", "answerBytes": 268, "answerChars": 268, "exitCode": 0, "usage": { "promptTokens": 70, "completionTokens": 67, "totalTokens": 137, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-random.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal values = {\"a\", \"b\", \"c\", \"d\"}\nlocal randomValue = values[math.random(#values)]\n```\nReason: For a sequential table, direct numeric indexing with `math.random(#values)` avoids the extra work `table.Random` does to handle non-sequential keys.", "durationMs": 4443, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f3e090701953d4a9", "startedAt": "2026-07-10T17:25:01.445Z", "completedAt": "2026-07-10T17:25:05.890Z", "answerBytes": 252, "answerChars": 252, "exitCode": 0, "usage": { "promptTokens": 66, "completionTokens": 63, "totalTokens": 129, "source": "estimated" } }, { "fixtureId": "gmod.perf.x-times-x.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal y = x * x\n```\nReason: `x * x` avoids function-call overhead and is the preferred fast form in a tight loop.", "durationMs": 3559, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "79d48008f6b5174f", "startedAt": "2026-07-10T17:25:01.854Z", "completedAt": "2026-07-10T17:25:05.414Z", "answerBytes": 120, "answerChars": 120, "exitCode": 0, "usage": { "promptTokens": 58, "completionTokens": 30, "totalTokens": 88, "source": "estimated" } }, { "fixtureId": "gmod.player-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attemptIndex": 1, "status": "pass", "detail": "Uses player.Iterator() (wiki-recommended Lua-cached iterator).", "finalResponse": "```lua\nfor _, ply in player.Iterator() do\n -- ...\nend\n```\nReason: `player.Iterator()` visits currently connected players without materializing `player.GetAll()` first.", "durationMs": 3690, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 2, "rubricVersion": "4", "promptHash": "fe3d7469535a0ca2", "startedAt": "2026-07-10T17:25:02.128Z", "completedAt": "2026-07-10T17:25:05.820Z", "answerBytes": 170, "answerChars": 170, "exitCode": 0, "usage": { "promptTokens": 80, "completionTokens": 43, "totalTokens": 123, "source": "estimated" } }, { "fixtureId": "gmod.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor ent in ents.Iterator() do\n if ent:GetClass() == \"prop_physics\" then\n -- use ent here\n end\nend\n```\nReason: `ents.Iterator()` streams entities one at a time, so it avoids building a full entity table before filtering for `prop_physics`.", "durationMs": 6968, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "3990c92253db85bb", "startedAt": "2026-07-10T17:25:02.268Z", "completedAt": "2026-07-10T17:25:09.238Z", "answerBytes": 258, "answerChars": 258, "exitCode": 0, "usage": { "promptTokens": 72, "completionTokens": 65, "totalTokens": 137, "source": "estimated" } }, { "fixtureId": "gmod.hook-add.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Registers a named hook with hook.Add.", "finalResponse": "```lua\nhook.Add(\"Think\", \"MyAddon.Think\", function()\n -- per-frame logic\nend)\n```\nReason: `hook.Add` registers a callback on Garry's Mod's `Think` hook using `MyAddon.Think` as the unique identifier.", "durationMs": 4039, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "12d2222fdb7d6b90", "startedAt": "2026-07-10T17:25:03.986Z", "completedAt": "2026-07-10T17:25:08.027Z", "answerBytes": 202, "answerChars": 202, "exitCode": 0, "usage": { "promptTokens": 78, "completionTokens": 51, "totalTokens": 129, "source": "estimated" } }, { "fixtureId": "gmod.isvalid.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Uses IsValid for entity/object validity.", "finalResponse": "```lua\nif IsValid(ent) then\n ent:SomeMethod()\nend\n```\nReason: In Garry's Mod, `IsValid(ent)` is the standard guard because it returns false for `nil`, invalid, or removed entities, so method calls on `ent` are only made when the reference is still safe to use.", "durationMs": 4368, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "35527312e59dfc13", "startedAt": "2026-07-10T17:25:04.146Z", "completedAt": "2026-07-10T17:25:08.516Z", "answerBytes": 263, "answerChars": 263, "exitCode": 0, "usage": { "promptTokens": 74, "completionTokens": 66, "totalTokens": 140, "source": "estimated" } }, { "fixtureId": "gmod.net-receivers.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nnet.Receive(\"MyAddon.Sync\", function(len, ply)\n -- Handle the incoming message here.\nend)\n```\nReason: `net.Receive` registers a callback for the incoming `MyAddon.Sync` net message.", "durationMs": 3853, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "24d54ee8366753a6", "startedAt": "2026-07-10T17:25:04.428Z", "completedAt": "2026-07-10T17:25:08.283Z", "answerBytes": 191, "answerChars": 191, "exitCode": 0, "usage": { "promptTokens": 64, "completionTokens": 48, "totalTokens": 112, "source": "estimated" } }, { "fixtureId": "gmod.perf.angle-zero.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "incorrect", "detail": "Did not match a fixture answer pattern.", "finalResponse": "```lua\nlocal ZERO_ANGLE = Angle(0, 0, 0)\n\n-- hot path\nent:SetAngles(ZERO_ANGLE)\n```\nReason: Reusing one shared zeroed `Angle` avoids allocating and initializing a new userdata/table on every call, which is faster in hot paths.", "durationMs": 9109, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f6bc97bf98280d62", "startedAt": "2026-07-10T17:25:04.904Z", "completedAt": "2026-07-10T17:25:14.014Z", "answerBytes": 226, "answerChars": 226, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 57, "totalTokens": 128, "source": "estimated" } }, { "fixtureId": "gmod.perf.config-var.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal MYADDON_COLOR = Color(255, 255, 255)\nlocal color = MYADDON_COLOR\n```\nReason: A flat local is faster to read because it avoids repeated table lookups and benefits from Lua's cheapest variable access path.", "durationMs": 4377, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "bcc4ff8eb162b699", "startedAt": "2026-07-10T17:25:05.074Z", "completedAt": "2026-07-10T17:25:09.453Z", "answerBytes": 216, "answerChars": 216, "exitCode": 0, "usage": { "promptTokens": 82, "completionTokens": 54, "totalTokens": 136, "source": "estimated" } }, { "fixtureId": "gmod.perf.darkrpvar.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal jobName = ply:getDarkRPVar(\"job\")\n\n-- instead of:\n-- local jobName = ply:GetNWString(\"job\")\n```\nReason: `ply:getDarkRPVar(\"job\")` is DarkRP's supported API for the synced job name, while `GetNWString(\"job\")` reaches into the underlying network storage directly and is less correct and less maintainable.", "durationMs": 6193, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "02a37daa699cded3", "startedAt": "2026-07-10T17:25:05.167Z", "completedAt": "2026-07-10T17:25:11.362Z", "answerBytes": 316, "answerChars": 316, "exitCode": 0, "usage": { "promptTokens": 66, "completionTokens": 79, "totalTokens": 145, "source": "estimated" } }, { "fixtureId": "gmod.perf.disttosqr.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nif pos1:DistToSqr(pos2) <= (maxDistance * maxDistance) then\n -- within range\nend\n```\nReason: `Vector:DistToSqr` avoids the square root in `Distance`, so it is the preferred method for threshold-only distance checks.", "durationMs": 5202, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "a5772470c6859cdc", "startedAt": "2026-07-10T17:25:05.298Z", "completedAt": "2026-07-10T17:25:10.501Z", "answerBytes": 225, "answerChars": 225, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 57, "totalTokens": 128, "source": "estimated" } }, { "fixtureId": "gmod.perf.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor ent in ents.Iterator() do\n -- visit every entity\nend\n```\nReason: `ents.Iterator()` is the wiki-recommended efficient choice for visiting every entity because it avoids building the table that `ents.GetAll()` returns.", "durationMs": 9612, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "306d3f1d05f28246", "startedAt": "2026-07-10T17:25:05.415Z", "completedAt": "2026-07-10T17:25:15.028Z", "answerBytes": 230, "answerChars": 230, "exitCode": 0, "usage": { "promptTokens": 66, "completionTokens": 58, "totalTokens": 124, "source": "estimated" } }, { "fixtureId": "gmod.perf.find-ents-near.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain 0 through 0 candidate loops.", "finalResponse": "```lua\nlocal radius = 512\nlocal nearby = ents.FindInSphere(ply:GetPos(), radius)\n\nfor i = 1, #nearby do\n local ent = nearby[i]\n if ent ~= ply then\n -- use ent\n end\nend\n```\nReason: `ents.FindInSphere` is a common fast built-in for \"near player\" spatial queries, avoids scanning all entities manually, and fits radial proximity better than making `FindInPVS` the default choice.", "durationMs": 6207, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "07c583fc3d9fe09f", "startedAt": "2026-07-10T17:25:05.486Z", "completedAt": "2026-07-10T17:25:11.695Z", "answerBytes": 392, "answerChars": 392, "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14910, "completionTokens": 212, "totalTokens": 15122, "reasoningTokens": 87, "cachedTokens": 14336 } }, { "fixtureId": "gmod.perf.find-players-near.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal function FindPlayersInRadius(pos, radius)\n local radiusSqr = radius * radius\n local out = {}\n\n for _, ply in ipairs(player.GetAll()) do\n if ply:GetPos():DistToSqr(pos) <= radiusSqr then\n out[#out + 1] = ply\n end\n end\n\n return out\nend\n```\nReason: `player.GetAll()` only iterates players, and `DistToSqr` avoids the square root work of full distance checks, so this skips scanning non-player entities and reduces per-call cost.", "durationMs": 7634, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "d139ba2e8ee38a2f", "startedAt": "2026-07-10T17:25:05.821Z", "completedAt": "2026-07-10T17:25:13.456Z", "answerBytes": 478, "answerChars": 478, "exitCode": 0, "usage": { "promptTokens": 79, "completionTokens": 120, "totalTokens": 199, "source": "estimated" } }, { "fixtureId": "gmod.perf.for-vs-while.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor i = 1, n do\n -- tight counting work\nend\n```\nReason: Numeric `for` is generally faster for `1..n` counting in Garry's Mod Lua because the loop control is handled by the VM with less per-iteration overhead than a manual `while` increment/check.", "durationMs": 20007, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "b3401a190a8aecf0", "startedAt": "2026-07-10T17:25:05.890Z", "completedAt": "2026-07-10T17:25:25.900Z", "answerBytes": 256, "answerChars": 256, "exitCode": 0, "usage": { "promptTokens": 65, "completionTokens": 64, "totalTokens": 129, "source": "estimated" } }, { "fixtureId": "gmod.perf.hook-once.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal players = player.GetHumans()\n\n-- Faster: one hook dispatch per tick, handlers loop players internally if needed.\nhook.Run(\"MyPlayersTick\", players)\n\n-- Slower:\n-- for i = 1, #players do\n-- hook.Run(\"MyPlayerTick\", players[i])\n-- end\n```\nReason: One `hook.Run` avoids repeated global hook lookup, vararg packing, and dispatcher overhead every tick; the per-player loop multiplies that overhead by player count.", "durationMs": 4809, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "ed1d3e09249fc8bb", "startedAt": "2026-07-10T17:25:06.117Z", "completedAt": "2026-07-10T17:25:10.930Z", "answerBytes": 426, "answerChars": 426, "exitCode": 0, "usage": { "promptTokens": 84, "completionTokens": 107, "totalTokens": 191, "source": "estimated" } }, { "fixtureId": "gmod.perf.hudpaint-cache.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal hp = LocalPlayer():Health()\n\ndraw.SimpleText(hp, \"DermaDefault\", x1, y1, color_white)\ndraw.SimpleText(hp .. \"%\", \"DermaDefault\", x2, y2, color_white)\ndraw.RoundedBox(4, x3, y3, hp * 2, 16, Color(200, 60, 60))\n```\nReason: Cache `LocalPlayer():Health()` once per frame and reuse it to avoid three repeated function calls in the hot `HUDPaint` path.", "durationMs": 5139, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3def450f3158ceff", "startedAt": "2026-07-10T17:25:06.117Z", "completedAt": "2026-07-10T17:25:11.264Z", "answerBytes": 359, "answerChars": 359, "exitCode": 0, "usage": { "promptTokens": 76, "completionTokens": 90, "totalTokens": 166, "source": "estimated" } }, { "fixtureId": "gmod.perf.local-color.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal COLOR_WHITE = Color(255, 255, 255)\n\nfunction PANEL:Paint(w, h)\n for i = 1, self.ItemCount do\n surface.SetDrawColor(COLOR_WHITE)\n -- draw work here\n end\nend\n```\nReason: Reuse one Color object outside the paint loop so the client stops allocating a new table every iteration.", "durationMs": 4971, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "16a7bed1b931d467", "startedAt": "2026-07-10T17:25:06.199Z", "completedAt": "2026-07-10T17:25:11.171Z", "answerBytes": 306, "answerChars": 306, "exitCode": 0, "usage": { "promptTokens": 63, "completionTokens": 77, "totalTokens": 140, "source": "estimated" } }, { "fixtureId": "gmod.perf.local-vs-global.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal sin = math.sin\n\nfor i = 1, n do\n out[i] = sin(i * step)\nend\n```\nReason: Caching `math.sin` in a local avoids repeated table lookups in the hot loop, which is a small but common Lua micro-optimization.", "durationMs": 3956, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "c116d8c651190647", "startedAt": "2026-07-10T17:25:06.749Z", "completedAt": "2026-07-10T17:25:10.708Z", "answerBytes": 216, "answerChars": 216, "exitCode": 0, "usage": { "promptTokens": 62, "completionTokens": 54, "totalTokens": 116, "source": "estimated" } }, { "fixtureId": "gmod.perf.localplayer-cache.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "partial", "detail": "Matched a partial-answer pattern.", "finalResponse": "```lua\nlocal localPlayer\n\nhook.Add(\"InitPostEntity\", \"CacheLocalPlayer\", function()\n localPlayer = LocalPlayer()\nend)\n\nhook.Add(\"Think\", \"UseCachedLocalPlayer\", function()\n local ply = localPlayer\n if not IsValid(ply) then\n ply = LocalPlayer()\n if not IsValid(ply) then return end\n localPlayer = ply\n end\n\n -- use ply here instead of calling LocalPlayer() repeatedly\n if ply:Alive() then\n -- hot-path logic\n end\nend)\n```\nReason: Cache the `LocalPlayer()` result once and reuse the local reference in hot paths, with a validity refresh only when needed, so you avoid repeated global lookup/function call overhead.", "durationMs": 8880, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "7398105af47b079e", "startedAt": "2026-07-10T17:25:07.411Z", "completedAt": "2026-07-10T17:25:16.294Z", "answerBytes": 661, "answerChars": 661, "exitCode": 0, "usage": { "promptTokens": 70, "completionTokens": 166, "totalTokens": 236, "source": "estimated" } }, { "fixtureId": "gmod.perf.looking-at.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "partial", "detail": "Matched a partial-answer pattern.", "finalResponse": "```lua\nlocal COS_FACING_EACH_OTHER = -0.85 -- about 148 degrees apart; closer to -1 means more directly facing each other\n\nlocal function ArePlayersRoughlyLookingAtEachOther(plyA, plyB)\n local aimA = plyA:GetAimVector()\n local aimB = plyB:GetAimVector()\n\n -- Unit vectors: dot product is cosine of the angle between them.\n -- If they are looking toward each other, their aim directions point opposite ways,\n -- so the dot product is near -1.\n return aimA:Dot(aimB) <= COS_FACING_EACH_OTHER\nend\n```\nReason: Using a dot-product threshold on normalized aim vectors avoids `math.acos`, giving the same facing test much cheaper and with less temporary math.", "durationMs": 6100, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "20b22f678a4b08d7", "startedAt": "2026-07-10T17:25:07.507Z", "completedAt": "2026-07-10T17:25:13.609Z", "answerBytes": 670, "answerChars": 670, "exitCode": 0, "usage": { "promptTokens": 91, "completionTokens": 168, "totalTokens": 259, "source": "estimated" } }, { "fixtureId": "gmod.perf.meta-vs-arg.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nSomeHelper(ply)\n```\nReason: For helpers you control, the direct function call is generally a bit faster because `:` adds implicit method lookup/self passing overhead, and it also avoids pretending the helper is part of the player metatable when it is not.", "durationMs": 4037, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "de332f7356830dfa", "startedAt": "2026-07-10T17:25:08.027Z", "completedAt": "2026-07-10T17:25:12.066Z", "answerBytes": 262, "answerChars": 262, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 66, "totalTokens": 137, "source": "estimated" } }, { "fixtureId": "gmod.perf.msg-vs-print.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nMsgN(\"hot path debug: \", someValue)\n```\nReason: `MsgN` writes directly to the console with less overhead than `print`, so it is the usual faster choice for hot-path debug output in Garry's Mod Lua.", "durationMs": 5093, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "6cbb4e4b2f19dd24", "startedAt": "2026-07-10T17:25:08.283Z", "completedAt": "2026-07-10T17:25:13.378Z", "answerBytes": 204, "answerChars": 204, "exitCode": 0, "usage": { "promptTokens": 63, "completionTokens": 51, "totalTokens": 114, "source": "estimated" } }, { "fixtureId": "gmod.perf.pairs-ipairs-for.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor i = 1, #nums do\n local n = nums[i]\n -- use n\nend\n```\nReason: `for i = 1, #t do` is typically fastest for a large dense numeric array because it avoids the extra iterator overhead of `pairs` and `ipairs` and does direct indexed access.", "durationMs": 11043, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "de5e4875fcf3e600", "startedAt": "2026-07-10T17:25:08.516Z", "completedAt": "2026-07-10T17:25:19.560Z", "answerBytes": 251, "answerChars": 251, "exitCode": 0, "usage": { "promptTokens": 73, "completionTokens": 63, "totalTokens": 136, "source": "estimated" } }, { "fixtureId": "gmod.perf.plain-lua-vs-fn.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\n-- Faster than chaining DarkRP fn combinators for hot paths: one tight loop, no closures, no intermediate lists.\nlocal players = player.GetAll()\nlocal out = {}\nlocal n = 0\n\nfor i = 1, #players do\n local ply = players[i]\n if IsValid(ply) and ply:Alive() and ply:isCP() then\n n = n + 1\n out[n] = ply:Nick()\n end\nend\n\nreturn out\n```\nReason: A single indexed loop avoids the extra function calls, closure dispatch, and intermediate table allocations that `fn.Filter(...):Map(...)` style pipelines add in plain Lua hot paths.", "durationMs": 13562, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "ead0a98b3ff6a388", "startedAt": "2026-07-10T17:25:09.238Z", "completedAt": "2026-07-10T17:25:22.802Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14912, "completionTokens": 228, "totalTokens": 15140, "reasoningTokens": 60, "cachedTokens": 0 }, "answerBytes": 550, "answerChars": 550 }, { "fixtureId": "gmod.perf.player-cache-hooks.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal activePlayers = player.GetAll()\n\nlocal function RefreshActivePlayers()\n activePlayers = player.GetAll()\nend\n\nhook.Add(\"PlayerInitialSpawn\", \"MyAddon.RefreshActivePlayers\", function()\n RefreshActivePlayers()\nend)\n\nhook.Add(\"PlayerDisconnected\", \"MyAddon.RefreshActivePlayers\", function()\n RefreshActivePlayers()\nend)\n\nhook.Add(\"Tick\", \"MyAddon.DoWork\", function()\n for i = 1, #activePlayers do\n local ply = activePlayers[i]\n if IsValid(ply) then\n -- per-player work here\n end\n end\nend)\n```\nReason: This avoids allocating and traversing a fresh `player.GetAll()` result every tick, and only rebuilds the player list when membership changes.", "durationMs": 13476, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "a834817afe9e6c25", "startedAt": "2026-07-10T17:25:09.454Z", "completedAt": "2026-07-10T17:25:22.932Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14907, "completionTokens": 207, "totalTokens": 15114, "reasoningTokens": 13, "cachedTokens": 14336 }, "answerBytes": 698, "answerChars": 698 }, { "fixtureId": "gmod.perf.ply-table-index.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "incorrect", "detail": "Did not match a fixture answer pattern.", "finalResponse": "```lua\nlocal healthCache = {}\n\nlocal function SetCachedHealth(ply, value)\n healthCache[ply:EntIndex()] = value\nend\n\nlocal function GetCachedHealth(ply)\n return healthCache[ply:EntIndex()]\nend\n```\nReason: Numeric array-style lookup by `EntIndex()` avoids repeated player table field/hash access and is typically faster for hot read paths than `ply.someField`.", "durationMs": 14150, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f18cc80df8999147", "startedAt": "2026-07-10T17:25:10.503Z", "completedAt": "2026-07-10T17:25:24.654Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14894, "completionTokens": 180, "totalTokens": 15074, "reasoningTokens": 69, "cachedTokens": 14336 }, "answerBytes": 364, "answerChars": 364 }, { "fixtureId": "gmod.perf.seq-tables.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal values = {\n \"alpha\",\n \"beta\",\n \"gamma\",\n \"delta\",\n}\n\nfor i = 1, #values do\n local value = values[i]\n -- use value\nend\n```\nReason: Sequential number-indexed arrays are faster to iterate because they use Lua's compact array part and avoid string-key hash lookups.", "durationMs": 12690, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "4778b0220c3f91a3", "startedAt": "2026-07-10T17:25:10.552Z", "completedAt": "2026-07-10T17:25:23.244Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14901, "completionTokens": 141, "totalTokens": 15042, "reasoningTokens": 39, "cachedTokens": 0 }, "answerBytes": 292, "answerChars": 292 }, { "fixtureId": "gmod.perf.setdrawcolor-split.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nsurface.SetDrawColor(255, 255, 255, 255)\n```\nReason: Passing raw numeric channels avoids the extra table/object access from a `Color` value and is the faster style in a hot draw path.", "durationMs": 11440, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "d110fde8a37bbe59", "startedAt": "2026-07-10T17:25:10.708Z", "completedAt": "2026-07-10T17:25:22.149Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14900, "completionTokens": 90, "totalTokens": 14990, "reasoningTokens": 22, "cachedTokens": 14336 }, "answerBytes": 190, "answerChars": 190 }, { "fixtureId": "gmod.perf.sqlite-vs-mysql.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal rows = sql.Query(\"SELECT money FROM players WHERE steamid = \" .. sql.SQLStr(steamID))\n```\nReason: Local SQLite via Garry's Mod `sql` library is typically lower latency on a single server because it avoids network round trips and remote database overhead.", "durationMs": 11635, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "4a3f5d2a20d71ac9", "startedAt": "2026-07-10T17:25:10.931Z", "completedAt": "2026-07-10T17:25:22.568Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14922, "completionTokens": 117, "totalTokens": 15039, "reasoningTokens": 38, "cachedTokens": 14336 }, "answerBytes": 267, "answerChars": 267 }, { "fixtureId": "gmod.perf.steamid-cache.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal steamIdCache = setmetatable({}, { __mode = \"k\" })\n\nlocal function GetSteamIDCached(ply)\n local steamId = steamIdCache[ply]\n if steamId ~= nil then\n return steamId\n end\n\n steamId = ply:SteamID()\n steamIdCache[ply] = steamId\n return steamId\nend\n\n-- hot path\nlocal function DoThing(players)\n for i = 1, #players do\n local ply = players[i]\n local steamId = GetSteamIDCached(ply)\n -- use steamId here\n end\nend\n\n-- optional invalidation on disconnect if you want explicit cleanup\nhook.Add(\"PlayerDisconnected\", \"ClearSteamIDCache\", function(ply)\n steamIdCache[ply] = nil\nend)\n```\nReason: This avoids repeated Lua-to-engine method calls on the hot path by computing each player's SteamID once and reusing it, while weak keys prevent stale player references from being kept alive.", "durationMs": 14399, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "e8a94aa454154982", "startedAt": "2026-07-10T17:25:11.172Z", "completedAt": "2026-07-10T17:25:25.572Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14891, "completionTokens": 267, "totalTokens": 15158, "reasoningTokens": 21, "cachedTokens": 0 }, "answerBytes": 840, "answerChars": 840 }, { "fixtureId": "gmod.perf.string-table-concat.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain 0 through 0 candidate loops.", "finalResponse": "```lua\nlocal pieces = {}\nfor i = 1, 10000 do\n pieces[i] = \"part\" .. i\nend\n\nlocal result = table.concat(pieces)\nprint(result)\n```\nReason: `table.concat` avoids creating a new intermediate string on every append, so it is much faster and uses less memory than repeated `..` concatenation in a loop.", "durationMs": 11478, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "b0cf9e04ae7c5dd9", "startedAt": "2026-07-10T17:25:11.265Z", "completedAt": "2026-07-10T17:25:22.745Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14892, "completionTokens": 122, "totalTokens": 15014, "reasoningTokens": 21, "cachedTokens": 14336 }, "answerBytes": 299, "answerChars": 299 }, { "fixtureId": "gmod.perf.surface-drawrect.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nsurface.SetDrawColor(r, g, b, a)\nsurface.DrawRect(x, y, w, h)\n```\nReason: `surface.DrawRect` avoids the extra Lua-side wrapper and corner setup work inside `draw.RoundedBox(0, ...)`, so it is slightly cheaper for plain rectangles.", "durationMs": 11924, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "696e8de769d9ad0e", "startedAt": "2026-07-10T17:25:11.363Z", "completedAt": "2026-07-10T17:25:23.289Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14896, "completionTokens": 131, "totalTokens": 15027, "reasoningTokens": 47, "cachedTokens": 14336 }, "answerBytes": 237, "answerChars": 237 }, { "fixtureId": "gmod.perf.surface-text.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal font = \"Trebuchet24\"\nlocal x, y = 32, 32\nlocal colorWhite = Color(255, 255, 255)\nlocal text = \"FPS: 144\"\n\nsurface.SetFont(font)\n\nhook.Add(\"HUDPaint\", \"FastSingleLineText\", function()\n surface.SetTextColor(colorWhite)\n surface.SetTextPos(x, y)\n surface.DrawText(text)\nend)\n```\nReason: `surface.SetTextPos` + `surface.DrawText` is the leanest common path for a single-line string because it avoids the extra alignment, vararg, and outline work in `draw.DrawText` and especially `draw.SimpleTextOutlined`. ", "durationMs": 14999, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "95d9fe4918ee4197", "startedAt": "2026-07-10T17:25:11.696Z", "completedAt": "2026-07-10T17:25:26.697Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14899, "completionTokens": 265, "totalTokens": 15164, "reasoningTokens": 93, "cachedTokens": 14336 }, "answerBytes": 525, "answerChars": 525 }, { "fixtureId": "gmod.perf.table-count-hash.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal len = #tbl\n```\nReason: `#` is the preferred built-in length operator for sequential array-like tables in Garry's Mod Lua.", "durationMs": 13678, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3e9bbfc8452e05e3", "startedAt": "2026-07-10T17:25:12.066Z", "completedAt": "2026-07-10T17:25:25.746Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14887, "completionTokens": 50, "totalTokens": 14937, "reasoningTokens": 0, "cachedTokens": 14336 }, "answerBytes": 134, "answerChars": 134 }, { "fixtureId": "gmod.perf.table-empty.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal t = {}\n\n-- When you only need to discard the old contents, this is generally faster:\nt = {}\n```\nReason: Rebinding `t` to a new table avoids clearing every existing entry one by one, so it is usually faster for large sequential tables unless other references must keep pointing at the same table.", "durationMs": 12048, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3a6344977f5bd475", "startedAt": "2026-07-10T17:25:12.317Z", "completedAt": "2026-07-10T17:25:24.366Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14894, "completionTokens": 151, "totalTokens": 15045, "reasoningTokens": 63, "cachedTokens": 14336 }, "answerBytes": 308, "answerChars": 308 }, { "fixtureId": "gmod.perf.table-hasvalue.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal allowed = {\n apple = true,\n banana = true,\n orange = true,\n}\n\nif allowed[value] then\n -- value exists in the set\nend\n```\nReason: Key lookup in a set-style table is O(1) on average, while `table.HasValue` scans a list linearly with O(n) cost.", "durationMs": 11622, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "7d064428a873fdfc", "startedAt": "2026-07-10T17:25:13.379Z", "completedAt": "2026-07-10T17:25:25.002Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14898, "completionTokens": 122, "totalTokens": 15020, "reasoningTokens": 24, "cachedTokens": 14336 }, "answerBytes": 266, "answerChars": 266 }, { "fixtureId": "gmod.perf.table-literal.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal t = { x = 1, y = 2, z = 3 }\n```\nReason: For a small table with known fields, the table constructor is typically faster and cleaner because it creates and populates the table in one step instead of doing extra field assignments after `{}`.", "durationMs": 11911, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f95a8fc610d720ad", "startedAt": "2026-07-10T17:25:13.457Z", "completedAt": "2026-07-10T17:25:25.369Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14895, "completionTokens": 111, "totalTokens": 15006, "reasoningTokens": 30, "cachedTokens": 14336 }, "answerBytes": 251, "answerChars": 251 }, { "fixtureId": "gmod.perf.table-random.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal value = seq[math.random(#seq)]\n```\nReason: For a sequential array, direct indexing with a random integer avoids the extra key traversal and non-array handling overhead in `table.Random`. ", "durationMs": 11495, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f3e090701953d4a9", "startedAt": "2026-07-10T17:25:13.610Z", "completedAt": "2026-07-10T17:25:25.106Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14893, "completionTokens": 103, "totalTokens": 14996, "reasoningTokens": 41, "cachedTokens": 14336 }, "answerBytes": 200, "answerChars": 200 }, { "fixtureId": "gmod.perf.x-times-x.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal xx = x * x\n```\nReason: `x * x` avoids the exponent operator's extra overhead and is the usual fastest form in a tight Garry's Mod Lua loop.", "durationMs": 11290, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "79d48008f6b5174f", "startedAt": "2026-07-10T17:25:14.015Z", "completedAt": "2026-07-10T17:25:25.306Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14886, "completionTokens": 98, "totalTokens": 14984, "reasoningTokens": 39, "cachedTokens": 14336 }, "answerBytes": 152, "answerChars": 152 }, { "fixtureId": "gmod.player-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attemptIndex": 1, "status": "pass", "detail": "Uses player.Iterator() (wiki-recommended Lua-cached iterator).", "finalResponse": "```lua\nfor _, ply in player.Iterator() do\n -- ...\nend\n```\nReason: `player.Iterator()` iterates connected players directly without first allocating the table returned by `player.GetAll()`. ", "durationMs": 11162, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 2, "rubricVersion": "4", "promptHash": "fe3d7469535a0ca2", "startedAt": "2026-07-10T17:25:15.029Z", "completedAt": "2026-07-10T17:25:26.192Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14904, "completionTokens": 82, "totalTokens": 14986, "reasoningTokens": 20, "cachedTokens": 14336 }, "answerBytes": 191, "answerChars": 191 }, { "fixtureId": "gmod.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor _, ent in ents.Iterator() do\n if ent:GetClass() == \"prop_physics\" then\n -- do something with ent\n end\nend\n```\nReason: `ents.Iterator()` walks entities lazily, so this filters `prop_physics` without first allocating a table of all map entities.", "durationMs": 20212, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "3990c92253db85bb", "startedAt": "2026-07-10T17:25:16.295Z", "completedAt": "2026-07-10T17:25:36.508Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14895, "completionTokens": 602, "totalTokens": 15497, "reasoningTokens": 516, "cachedTokens": 0 }, "answerBytes": 267, "answerChars": 267 }, { "fixtureId": "gmod.hook-add.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Registers a named hook with hook.Add.", "finalResponse": "```lua\nhook.Add(\"Think\", \"MyAddon.Think\", function()\n -- runs every Think\nend)\n```\nReason: `hook.Add` registers a callback on the `Think` hook, and `MyAddon.Think` is the unique identifier for this handler.", "durationMs": 6908, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "12d2222fdb7d6b90", "startedAt": "2026-07-10T17:25:19.561Z", "completedAt": "2026-07-10T17:25:26.471Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14897, "completionTokens": 155, "totalTokens": 15052, "reasoningTokens": 79, "cachedTokens": 0 }, "answerBytes": 209, "answerChars": 209 }, { "fixtureId": "gmod.isvalid.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Uses IsValid for entity/object validity.", "finalResponse": "```lua\nif IsValid(ent) then\n ent:SomeMethod()\nend\n```\nReason: `IsValid(ent)` safely rejects `nil`, invalid entities, and NULL entity references before you call methods on `ent`.", "durationMs": 23815, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "35527312e59dfc13", "startedAt": "2026-07-10T17:25:22.150Z", "completedAt": "2026-07-10T17:25:45.967Z", "answerBytes": 180, "answerChars": 180, "exitCode": 0, "usage": { "promptTokens": 74, "completionTokens": 45, "totalTokens": 119, "source": "estimated" } }, { "fixtureId": "gmod.net-receivers.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nnet.Receive(\"MyAddon.Sync\", function(len, ply)\n print(\"Received MyAddon.Sync\")\nend)\n```\nReason: `net.Receive` registers a callback that runs whenever the `MyAddon.Sync` net message arrives.", "durationMs": 5364, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "24d54ee8366753a6", "startedAt": "2026-07-10T17:25:22.569Z", "completedAt": "2026-07-10T17:25:27.934Z", "answerBytes": 199, "answerChars": 199, "exitCode": 0, "usage": { "promptTokens": 64, "completionTokens": 50, "totalTokens": 114, "source": "estimated" } }, { "fixtureId": "gmod.perf.angle-zero.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "incorrect", "detail": "Did not match a fixture answer pattern.", "finalResponse": "```lua\n-- Slow in a hot path: allocates a new Angle every time.\nlocal ang1 = Angle(0, 0, 0)\n\n-- Faster: reuse the engine-provided shared zero angle.\nlocal ang2 = angle_zero\n```\nReason: `angle_zero` avoids repeated `Angle(0, 0, 0)` allocations and constructor work; only use it when you will not mutate the shared angle.", "durationMs": 9712, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f6bc97bf98280d62", "startedAt": "2026-07-10T17:25:22.746Z", "completedAt": "2026-07-10T17:25:32.459Z", "answerBytes": 319, "answerChars": 319, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 80, "totalTokens": 151, "source": "estimated" } }, { "fixtureId": "gmod.perf.config-var.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal MYADDON_COLOR = Color(255, 0, 0)\n\nlocal color = MYADDON_COLOR\n```\nReason: A local variable is fastest for hot reads because it avoids repeated table indexing (`myaddon.config.color`) and global lookup overhead.", "durationMs": 8988, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "bcc4ff8eb162b699", "startedAt": "2026-07-10T17:25:22.803Z", "completedAt": "2026-07-10T17:25:31.792Z", "answerBytes": 223, "answerChars": 223, "exitCode": 0, "usage": { "promptTokens": 82, "completionTokens": 56, "totalTokens": 138, "source": "estimated" } }, { "fixtureId": "gmod.perf.darkrpvar.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal jobName = ply:getDarkRPVar(\"job\")\n\n-- Not preferred:\n-- local jobName = ply:GetNWString(\"job\")\n```\nReason: `getDarkRPVar(\"job\")` is DarkRP’s intended API for the networked job name, while `GetNWString(\"job\")` bypasses the gamemode-specific contract.", "durationMs": 9945, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "02a37daa699cded3", "startedAt": "2026-07-10T17:25:22.933Z", "completedAt": "2026-07-10T17:25:32.880Z", "answerBytes": 264, "answerChars": 262, "exitCode": 0, "usage": { "promptTokens": 66, "completionTokens": 66, "totalTokens": 132, "source": "estimated" } }, { "fixtureId": "gmod.perf.disttosqr.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal maxDistance = 300\nif pos1:DistToSqr(pos2) <= maxDistance * maxDistance then\n -- within 300 units\nend\n```\nReason: `Vector:DistToSqr` avoids the square root from `Distance`, so it is the preferred faster check when you only need to compare against a distance threshold.", "durationMs": 15903, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "a5772470c6859cdc", "startedAt": "2026-07-10T17:25:23.246Z", "completedAt": "2026-07-10T17:25:39.153Z", "answerBytes": 283, "answerChars": 283, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 71, "totalTokens": 142, "source": "estimated" } }, { "fixtureId": "gmod.perf.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain 1 through 1 candidate loops.", "finalResponse": "```lua\nif ents.Iterator then\n for _, ent in ents.Iterator() do\n -- use ent\n end\nelse\n for _, ent in ipairs(ents.GetAll()) do\n -- use ent\n end\nend\n```\nReason: `ents.Iterator()` avoids building the full entity table and is the preferred efficient full-entity iterator; `ipairs(ents.GetAll())` is a compatible fallback.", "durationMs": 16747, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "306d3f1d05f28246", "startedAt": "2026-07-10T17:25:23.290Z", "completedAt": "2026-07-10T17:25:40.038Z", "answerBytes": 342, "answerChars": 342, "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14893, "completionTokens": 622, "totalTokens": 15515, "reasoningTokens": 516, "cachedTokens": 14336 } }, { "fixtureId": "gmod.perf.find-ents-near.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain 0 through 0 candidate loops.", "finalResponse": "```lua\nlocal radius = 512\nlocal nearby = ents.FindInSphere(ply:GetPos(), radius)\n\nfor i = 1, #nearby do\n local ent = nearby[i]\n if ent ~= ply then\n -- use ent\n end\nend\n```\nReason: `ents.FindInSphere` is a common fast default for nearby-entity queries because it does direct radius-based broad-phase lookup without the heavier visibility/path semantics of `FindInPVS`. ", "durationMs": 18378, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "07c583fc3d9fe09f", "startedAt": "2026-07-10T17:25:24.367Z", "completedAt": "2026-07-10T17:25:42.746Z", "answerBytes": 384, "answerChars": 384, "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14910, "completionTokens": 337, "totalTokens": 15247, "reasoningTokens": 217, "cachedTokens": 14336 } }, { "fixtureId": "gmod.perf.find-players-near.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal player_GetAll = player.GetAll\n\nlocal function FindPlayersNear(pos, radius, out)\n out = out or {}\n\n local radiusSqr = radius * radius\n local players = player_GetAll()\n local count = 0\n\n for i = 1, #players do\n local ply = players[i]\n if pos:DistToSqr(ply:GetPos()) <= radiusSqr then\n count = count + 1\n out[count] = ply\n end\n end\n\n for i = count + 1, #out do\n out[i] = nil\n end\n\n return out\nend\n```\nReason: It only scans players instead of every entity and uses squared-distance checks, so it avoids the extra entity work and sqrt cost from `ents.FindInSphere` plus filtering.", "durationMs": 15483, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "d139ba2e8ee38a2f", "startedAt": "2026-07-10T17:25:24.655Z", "completedAt": "2026-07-10T17:25:40.139Z", "answerBytes": 665, "answerChars": 665, "exitCode": 0, "usage": { "promptTokens": 79, "completionTokens": 167, "totalTokens": 246, "source": "estimated" } }, { "fixtureId": "gmod.perf.for-vs-while.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor i = 1, n do\n -- tight loop body\nend\n```\nReason: Numeric `for` is generally faster for 1..n counting in Lua because the loop control stays in the VM with less per-iteration overhead than a manual `while` increment/check.", "durationMs": 20153, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "b3401a190a8aecf0", "startedAt": "2026-07-10T17:25:25.003Z", "completedAt": "2026-07-10T17:25:45.159Z", "answerBytes": 233, "answerChars": 233, "exitCode": 0, "usage": { "promptTokens": 65, "completionTokens": 59, "totalTokens": 124, "source": "estimated" } }, { "fixtureId": "gmod.perf.hook-once.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nhook.Add(\"Think\", \"NotifyPlayersOnce\", function()\n local players = player.GetAll()\n\n -- Faster: one hook dispatch, handlers iterate players themselves.\n hook.Run(\"MyPlayersTick\", players)\n\n -- Slower:\n -- for i = 1, #players do\n -- hook.Run(\"MyPlayerTick\", players[i])\n -- end\nend)\n\nhook.Add(\"MyPlayersTick\", \"MyCustomLogic\", function(players)\n for i = 1, #players do\n local ply = players[i]\n -- custom per-player logic here\n end\nend)\n```\nReason: One `hook.Run` call avoids repeating hook dispatch, vararg setup, and listener iteration for every player, so total overhead is lower.", "durationMs": 34907, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "ed1d3e09249fc8bb", "startedAt": "2026-07-10T17:25:25.107Z", "completedAt": "2026-07-10T17:26:00.016Z", "answerBytes": 633, "answerChars": 633, "exitCode": 0, "usage": { "promptTokens": 84, "completionTokens": 159, "totalTokens": 243, "source": "estimated" } }, { "fixtureId": "gmod.perf.hudpaint-cache.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nhook.Add(\"HUDPaint\", \"DrawHealthFast\", function()\n local lp = LocalPlayer()\n local health = lp:Health()\n\n draw.SimpleText(health, \"DermaLarge\", 32, 32, color_white)\n draw.RoundedBox(4, 32, 72, health * 2, 20, Color(200, 50, 50))\n draw.SimpleText(\"HP: \" .. health, \"Trebuchet24\", 32, 100, color_white)\nend)\n```\nReason: Cache `LocalPlayer()` and `:Health()` once per frame to avoid repeating the same function calls three times inside `HUDPaint`.", "durationMs": 6950, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3def450f3158ceff", "startedAt": "2026-07-10T17:25:25.308Z", "completedAt": "2026-07-10T17:25:32.261Z", "answerBytes": 466, "answerChars": 466, "exitCode": 0, "usage": { "promptTokens": 76, "completionTokens": 117, "totalTokens": 193, "source": "estimated" } }, { "fixtureId": "gmod.perf.local-color.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal WHITE = Color(255, 255, 255)\n\nfunction PANEL:Paint(w, h)\n surface.SetDrawColor(WHITE)\n\n for i = 1, self.BarCount do\n surface.DrawRect((i - 1) * 8, 0, 6, h)\n end\nend\n```\nReason: `Color(...)` allocates/builds a color object each iteration; caching one immutable white color outside the paint loop removes that repeated work.", "durationMs": 6298, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "16a7bed1b931d467", "startedAt": "2026-07-10T17:25:25.370Z", "completedAt": "2026-07-10T17:25:31.669Z", "answerBytes": 351, "answerChars": 351, "exitCode": 0, "usage": { "promptTokens": 63, "completionTokens": 88, "totalTokens": 151, "source": "estimated" } }, { "fixtureId": "gmod.perf.local-vs-global.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal sin = math.sin\n\nfor i = 1, 100000 do\n local y = sin(i)\nend\n```\nReason: Caching `math.sin` in a local avoids repeated table lookups (`math.sin`) inside the hot loop, which is slightly faster in Lua.", "durationMs": 5437, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "c116d8c651190647", "startedAt": "2026-07-10T17:25:25.573Z", "completedAt": "2026-07-10T17:25:31.013Z", "answerBytes": 213, "answerChars": 213, "exitCode": 0, "usage": { "promptTokens": 62, "completionTokens": 54, "totalTokens": 116, "source": "estimated" } }, { "fixtureId": "gmod.perf.localplayer-cache.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "partial", "detail": "Matched a partial-answer pattern.", "finalResponse": "```lua\nlocal LocalPlayer = LocalPlayer\nlocal IsValid = IsValid\n\nlocal cachedLocalPlayer\n\nlocal function LP()\n if not IsValid(cachedLocalPlayer) then\n cachedLocalPlayer = LocalPlayer()\n end\n\n return cachedLocalPlayer\nend\n\nhook.Add(\"HUDPaint\", \"ExampleCachedLocalPlayer\", function()\n local ply = LP()\n if not IsValid(ply) then return end\n\n local health = ply:Health()\n local armor = ply:Armor()\n draw.SimpleText(health .. \" / \" .. armor, \"DermaDefault\", 16, 16, color_white)\nend)\n```\nReason: This caches both the global function lookup and the player entity, so repeated code uses a local reference and only refreshes it when the cached player is invalid.", "durationMs": 14663, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "7398105af47b079e", "startedAt": "2026-07-10T17:25:25.747Z", "completedAt": "2026-07-10T17:25:40.411Z", "answerBytes": 684, "answerChars": 684, "exitCode": 0, "usage": { "promptTokens": 70, "completionTokens": 171, "totalTokens": 241, "source": "estimated" } }, { "fixtureId": "gmod.perf.looking-at.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "partial", "detail": "Matched a partial-answer pattern.", "finalResponse": "```lua\nlocal LOOK_DOT = 0.8 -- about 37 degrees of slack\n\nlocal function RoughlyLookingAtEachOther(plyA, plyB)\n local toB = (plyB:EyePos() - plyA:EyePos()):GetNormalized()\n local toA = -toB\n\n return plyA:GetAimVector():Dot(toB) >= LOOK_DOT\n and plyB:GetAimVector():Dot(toA) >= LOOK_DOT\nend\n```\nReason: Dot products avoid `math.acos`, so this stays cheap while still checking that each player's aim vector points toward the other player's eye position within a tunable threshold.", "durationMs": 17802, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "20b22f678a4b08d7", "startedAt": "2026-07-10T17:25:25.901Z", "completedAt": "2026-07-10T17:25:43.704Z", "answerBytes": 493, "answerChars": 493, "exitCode": 0, "usage": { "promptTokens": 91, "completionTokens": 124, "totalTokens": 215, "source": "estimated" } }, { "fixtureId": "gmod.perf.meta-vs-arg.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nSomeHelper(ply)\n```\nReason: `SomeHelper(ply)` is generally the better choice for a helper you control because a local/direct function call avoids the per-call table method lookup behind `ply:SomeHelper()`, and `obj:method()` is mainly for real object methods, not free helpers.", "durationMs": 7569, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "de332f7356830dfa", "startedAt": "2026-07-10T17:25:26.193Z", "completedAt": "2026-07-10T17:25:33.764Z", "answerBytes": 284, "answerChars": 284, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 71, "totalTokens": 142, "source": "estimated" } }, { "fixtureId": "gmod.perf.msg-vs-print.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nMsgN(\"debug: hot path hit\")\n```\nReason: `MsgN` is typically faster than `print` in Garry's Mod for a simple debug line because it writes directly to the console without `print`'s extra argument formatting overhead.", "durationMs": 5478, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "6cbb4e4b2f19dd24", "startedAt": "2026-07-10T17:25:26.472Z", "completedAt": "2026-07-10T17:25:31.951Z", "answerBytes": 221, "answerChars": 221, "exitCode": 0, "usage": { "promptTokens": 63, "completionTokens": 56, "totalTokens": 119, "source": "estimated" } }, { "fixtureId": "gmod.perf.pairs-ipairs-for.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor i = 1, #numbers do\n local value = numbers[i]\n -- use value\nend\n```\nReason: `for i = 1, #numbers do` is the fastest typical choice for a large dense numeric array in Garry's Mod Lua because it avoids the iterator overhead of `pairs` and `ipairs` and does direct indexed access.", "durationMs": 7327, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "de5e4875fcf3e600", "startedAt": "2026-07-10T17:25:26.697Z", "completedAt": "2026-07-10T17:25:34.026Z", "answerBytes": 293, "answerChars": 293, "exitCode": 0, "usage": { "promptTokens": 73, "completionTokens": 74, "totalTokens": 147, "source": "estimated" } }, { "fixtureId": "gmod.perf.plain-lua-vs-fn.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\n-- Equivalent to a filter->map pipeline, but done in one pass.\nlocal function collectPoliceNames(players)\n local out = {}\n local outN = 0\n\n for i = 1, #players do\n local ply = players[i]\n if IsValid(ply) and ply:Alive() and ply:Team() == TEAM_POLICE then\n outN = outN + 1\n out[outN] = ply:Nick()\n end\n end\n\n return out\nend\n\nlocal policeNames = collectPoliceNames(player.GetAll())\n```\nReason: A single indexed loop is usually faster than DarkRP's `fn` filter/map chain because it avoids extra closures, function-call overhead, and intermediate tables while doing the same work in one pass.", "durationMs": 16115, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "ead0a98b3ff6a388", "startedAt": "2026-07-10T17:25:27.935Z", "completedAt": "2026-07-10T17:25:44.051Z", "answerBytes": 653, "answerChars": 653, "exitCode": 0, "usage": { "promptTokens": 76, "completionTokens": 164, "totalTokens": 240, "source": "estimated" } }, { "fixtureId": "gmod.perf.player-cache-hooks.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal cachedPlayers = player.GetAll()\n\nlocal function RefreshPlayerCache()\n cachedPlayers = player.GetAll()\nend\n\nhook.Add(\"PlayerInitialSpawn\", \"MyAddon.RefreshPlayerCache\", RefreshPlayerCache)\nhook.Add(\"PlayerDisconnected\", \"MyAddon.RefreshPlayerCache\", RefreshPlayerCache)\n\nhook.Add(\"Tick\", \"MyAddon.DoPerPlayerWork\", function()\n for i = 1, #cachedPlayers do\n local ply = cachedPlayers[i]\n if IsValid(ply) then\n -- per-player work here\n end\n end\nend)\n```\nReason: `player.GetAll()` allocates/builds a fresh table, so caching it and only rebuilding on join/leave removes that per-tick work from `Tick`/`Think`.", "durationMs": 9685, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "a834817afe9e6c25", "startedAt": "2026-07-10T17:25:31.014Z", "completedAt": "2026-07-10T17:25:40.701Z", "answerBytes": 658, "answerChars": 658, "exitCode": 0, "usage": { "promptTokens": 88, "completionTokens": 165, "totalTokens": 253, "source": "estimated" } }, { "fixtureId": "gmod.perf.ply-table-index.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "incorrect", "detail": "Did not match a fixture answer pattern.", "finalResponse": "```lua\nlocal tempByEntIndex = {}\n\nlocal function SetTempData(ply, value)\n tempByEntIndex[ply:EntIndex()] = value\nend\n\nlocal function GetTempData(ply)\n return tempByEntIndex[ply:EntIndex()]\nend\n\nhook.Add(\"PlayerDisconnected\", \"ClearTempData\", function(ply)\n tempByEntIndex[ply:EntIndex()] = nil\nend)\n```\nReason: A plain Lua table indexed by `EntIndex()` avoids per-read entity field/metatable lookup, so hot reads are cheaper; clearing on disconnect prevents slot reuse bugs.", "durationMs": 12273, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f18cc80df8999147", "startedAt": "2026-07-10T17:25:31.670Z", "completedAt": "2026-07-10T17:25:43.945Z", "answerBytes": 483, "answerChars": 483, "exitCode": 0, "usage": { "promptTokens": 69, "completionTokens": 121, "totalTokens": 190, "source": "estimated" } }, { "fixtureId": "gmod.perf.seq-tables.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal values = {\n \"alpha\",\n \"beta\",\n \"gamma\",\n \"delta\",\n}\n\nfor i = 1, #values do\n local value = values[i]\n -- use value\nend\n```\nReason: A sequential number-indexed array is faster to iterate than a string-keyed map because it uses dense integer keys and a simple numeric loop instead of hash-key traversal.", "durationMs": 8046, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "4778b0220c3f91a3", "startedAt": "2026-07-10T17:25:31.793Z", "completedAt": "2026-07-10T17:25:39.842Z", "answerBytes": 331, "answerChars": 331, "exitCode": 0, "usage": { "promptTokens": 79, "completionTokens": 83, "totalTokens": 162, "source": "estimated" } }, { "fixtureId": "gmod.perf.setdrawcolor-split.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nsurface.SetDrawColor(255, 255, 255, 255)\n```\nReason: Passing raw numbers is faster in a hot draw path because it avoids the extra Color/table argument handling that the `Color` overload requires.", "durationMs": 6530, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "d110fde8a37bbe59", "startedAt": "2026-07-10T17:25:31.952Z", "completedAt": "2026-07-10T17:25:38.484Z", "answerBytes": 202, "answerChars": 202, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 51, "totalTokens": 122, "source": "estimated" } }, { "fixtureId": "gmod.perf.sqlite-vs-mysql.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal rows = sql.Query(\"SELECT coins FROM players WHERE steamid = \" .. sql.SQLStr(ply:SteamID64()))\n```\nReason: Local SQLite via Garry's Mod `sql` library is typically lower latency on a single server because it avoids network round trips and remote DB connection overhead that MySQL adds.", "durationMs": 8669, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "4a3f5d2a20d71ac9", "startedAt": "2026-07-10T17:25:32.261Z", "completedAt": "2026-07-10T17:25:40.932Z", "answerBytes": 296, "answerChars": 296, "exitCode": 0, "usage": { "promptTokens": 100, "completionTokens": 74, "totalTokens": 174, "source": "estimated" } }, { "fixtureId": "gmod.perf.steamid-cache.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal SID_CACHE_KEY = \"__cached_steamid\"\n\nlocal function GetSteamIDCached(ply)\n local sid = ply[SID_CACHE_KEY]\n if sid ~= nil then\n return sid\n end\n\n sid = ply:SteamID()\n ply[SID_CACHE_KEY] = sid\n return sid\nend\n\n-- Hot path\nfor i = 1, #players do\n local ply = players[i]\n local steamID = GetSteamIDCached(ply)\n -- use steamID\nend\n```\nReason: `ply:SteamID()` crosses into engine code each call, while a cached Lua field turns repeated lookups for the same player into a cheap table access.", "durationMs": 14965, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "e8a94aa454154982", "startedAt": "2026-07-10T17:25:32.460Z", "completedAt": "2026-07-10T17:25:47.427Z", "answerBytes": 530, "answerChars": 530, "exitCode": 0, "usage": { "promptTokens": 63, "completionTokens": 133, "totalTokens": 196, "source": "estimated" } }, { "fixtureId": "gmod.perf.string-table-concat.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain 0 through 0 candidate loops.", "finalResponse": "```lua\nlocal pieces = {}\n\nfor i = 1, 100000 do\n pieces[i] = \"part\" .. i\nend\n\n-- Slow: repeated .. creates a new string every iteration.\nlocal slow = \"\"\nfor i = 1, #pieces do\n slow = slow .. pieces[i]\nend\n\n-- Fast: collect pieces, then join once.\nlocal fast = table.concat(pieces)\n```\nReason: `table.concat` is faster because it avoids reallocating and copying the whole accumulated string on every `..` append.", "durationMs": 6493, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "b0cf9e04ae7c5dd9", "startedAt": "2026-07-10T17:25:32.881Z", "completedAt": "2026-07-10T17:25:39.376Z", "answerBytes": 416, "answerChars": 416, "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14894, "completionTokens": 216, "totalTokens": 15110, "reasoningTokens": 76, "cachedTokens": 14336 } }, { "fixtureId": "gmod.perf.surface-drawrect.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nsurface.SetDrawColor(255, 255, 255, 255)\nsurface.DrawRect(x, y, w, h)\n```\nReason: `surface.DrawRect` is a more direct rectangle draw path, while `draw.RoundedBox(0, ...)` adds wrapper overhead you do not need for non-rounded boxes.", "durationMs": 7067, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "696e8de769d9ad0e", "startedAt": "2026-07-10T17:25:33.765Z", "completedAt": "2026-07-10T17:25:40.833Z", "answerBytes": 238, "answerChars": 238, "exitCode": 0, "usage": { "promptTokens": 69, "completionTokens": 60, "totalTokens": 129, "source": "estimated" } }, { "fixtureId": "gmod.perf.surface-text.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal FONT = \"DermaDefaultBold\"\nlocal TEXT = \"FPS\"\nlocal X, Y = 24, 24\nlocal R, G, B, A = 255, 255, 255, 255\n\nhook.Add(\"HUDPaint\", \"FastSingleLineText\", function()\n surface.SetFont(FONT)\n surface.SetTextColor(R, G, B, A)\n surface.SetTextPos(X, Y)\n surface.DrawText(TEXT)\nend)\n```\nReason: `surface.SetFont` + `surface.SetTextPos` + `surface.DrawText` is the leanest path here for a single-line string; `draw.DrawText` does extra layout/alignment work and `draw.SimpleTextOutlined` adds multiple outline passes.", "durationMs": 10952, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "95d9fe4918ee4197", "startedAt": "2026-07-10T17:25:34.027Z", "completedAt": "2026-07-10T17:25:44.980Z", "answerBytes": 528, "answerChars": 528, "exitCode": 0, "usage": { "promptTokens": 76, "completionTokens": 132, "totalTokens": 208, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-count-hash.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal len = #myTable\n```\nReason: `#` is the preferred built-in length operator for sequential array-like tables; it is direct and avoids `table.Count`, which is for counting all keys.", "durationMs": 4863, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3e9bbfc8452e05e3", "startedAt": "2026-07-10T17:25:36.509Z", "completedAt": "2026-07-10T17:25:41.373Z", "answerBytes": 190, "answerChars": 190, "exitCode": 0, "usage": { "promptTokens": 62, "completionTokens": 48, "totalTokens": 110, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-empty.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nt = {}\n```\nReason: `t = {}` is generally faster because it allocates a new table instead of clearing every existing element one by one like `table.Empty(t)`, as long as you do not need to preserve other references to the same table.", "durationMs": 7262, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3a6344977f5bd475", "startedAt": "2026-07-10T17:25:38.485Z", "completedAt": "2026-07-10T17:25:45.748Z", "answerBytes": 239, "answerChars": 239, "exitCode": 0, "usage": { "promptTokens": 73, "completionTokens": 60, "totalTokens": 133, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-hasvalue.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal allowed = {\n pistol = true,\n shotgun = true,\n crowbar = true,\n}\n\nif allowed[className] then\n -- value exists in the set\nend\n```\nReason: `table.HasValue` scans a list linearly, while a keyed lookup table gives near-constant-time membership checks.", "durationMs": 5560, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "7d064428a873fdfc", "startedAt": "2026-07-10T17:25:39.154Z", "completedAt": "2026-07-10T17:25:44.715Z", "answerBytes": 271, "answerChars": 271, "exitCode": 0, "usage": { "promptTokens": 67, "completionTokens": 68, "totalTokens": 135, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-literal.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\n-- Faster\nlocal t = {\n a = 1,\n b = 2,\n c = 3,\n}\n\n-- Slower\n-- local t = {}\n-- t.a = 1\n-- t.b = 2\n-- t.c = 3\n```\nReason: For a small table with known keys, the table constructor does the work in one allocation/initialization step and avoids separate field-set operations after `{}`.", "durationMs": 7375, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f95a8fc610d720ad", "startedAt": "2026-07-10T17:25:39.377Z", "completedAt": "2026-07-10T17:25:46.753Z", "answerBytes": 297, "answerChars": 297, "exitCode": 0, "usage": { "promptTokens": 70, "completionTokens": 75, "totalTokens": 145, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-random.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal function randomSequential(t)\n return t[math.random(#t)]\nend\n\nlocal value = randomSequential(mySequentialTable)\n```\nReason: For a dense 1..n array, direct indexing with `math.random(#t)` avoids the extra generic work `table.Random` does to support non-sequential tables.", "durationMs": 7403, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f3e090701953d4a9", "startedAt": "2026-07-10T17:25:39.843Z", "completedAt": "2026-07-10T17:25:47.247Z", "answerBytes": 285, "answerChars": 285, "exitCode": 0, "usage": { "promptTokens": 66, "completionTokens": 72, "totalTokens": 138, "source": "estimated" } }, { "fixtureId": "gmod.perf.x-times-x.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal squared = x * x\n```\nReason: `x * x` is the preferred fast form in a tight loop because it avoids function-call overhead and exponentiation dispatch.", "durationMs": 4487, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "79d48008f6b5174f", "startedAt": "2026-07-10T17:25:40.039Z", "completedAt": "2026-07-10T17:25:44.528Z", "answerBytes": 161, "answerChars": 161, "exitCode": 0, "usage": { "promptTokens": 58, "completionTokens": 41, "totalTokens": 99, "source": "estimated" } }, { "fixtureId": "gmod.player-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attemptIndex": 1, "status": "pass", "detail": "Uses player.Iterator() (wiki-recommended Lua-cached iterator).", "finalResponse": "```lua\nfor _, ply in player.Iterator() do\n -- ...\nend\n```\nReason: `player.Iterator()` visits every connected player without allocating the temporary table that `player.GetAll()` creates.", "durationMs": 10619, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 2, "rubricVersion": "4", "promptHash": "fe3d7469535a0ca2", "startedAt": "2026-07-10T17:25:40.143Z", "completedAt": "2026-07-10T17:25:50.764Z", "answerBytes": 189, "answerChars": 189, "exitCode": 0, "usage": { "promptTokens": 80, "completionTokens": 48, "totalTokens": 128, "source": "estimated" } }, { "fixtureId": "gmod.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor _, ent in ents.Iterator() do\n if ent:GetClass() == \"prop_physics\" then\n -- do something with ent\n end\nend\n```\nReason: `ents.Iterator()` streams entities one by one, so it avoids allocating a table of all entities before filtering by class.", "durationMs": 13332, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "3990c92253db85bb", "startedAt": "2026-07-10T17:25:40.412Z", "completedAt": "2026-07-10T17:25:53.746Z", "answerBytes": 263, "answerChars": 263, "exitCode": 0, "usage": { "promptTokens": 72, "completionTokens": 66, "totalTokens": 138, "source": "estimated" } }, { "fixtureId": "gmod.hook-add.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Registers a named hook with hook.Add.", "finalResponse": "```lua\nhook.Add(\"Think\", \"MyAddon.Think\", function()\n -- Think callback\nend)\n```\nReason: `hook.Add` registers a function on Garry's Mod's `Think` hook, and `MyAddon.Think` is the unique identifier for this callback.", "durationMs": 5455, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "12d2222fdb7d6b90", "startedAt": "2026-07-10T17:25:40.702Z", "completedAt": "2026-07-10T17:25:46.158Z", "answerBytes": 218, "answerChars": 218, "exitCode": 0, "usage": { "promptTokens": 78, "completionTokens": 55, "totalTokens": 133, "source": "estimated" } }, { "fixtureId": "gmod.isvalid.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Uses IsValid for entity/object validity.", "finalResponse": "```lua\nif IsValid(ent) then\n ent:SomeMethod()\nend\n```\nReason: `IsValid(ent)` is the standard Garry's Mod guard and returns false for `nil`, `NULL`, and removed/invalid entities, so method calls only run when the reference is still usable.", "durationMs": 6804, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "35527312e59dfc13", "startedAt": "2026-07-10T17:25:40.834Z", "completedAt": "2026-07-10T17:25:47.640Z", "answerBytes": 241, "answerChars": 241, "exitCode": 0, "usage": { "promptTokens": 74, "completionTokens": 61, "totalTokens": 135, "source": "estimated" } }, { "fixtureId": "gmod.net-receivers.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nnet.Receive(\"MyAddon.Sync\", function(len, ply)\n -- Handle the incoming MyAddon.Sync message here.\nend)\n```\nReason: `net.Receive` registers a callback that runs whenever the `MyAddon.Sync` net message arrives.", "durationMs": 11476, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "24d54ee8366753a6", "startedAt": "2026-07-10T17:25:40.933Z", "completedAt": "2026-07-10T17:25:52.410Z", "answerBytes": 218, "answerChars": 218, "exitCode": 0, "usage": { "promptTokens": 64, "completionTokens": 55, "totalTokens": 119, "source": "estimated" } }, { "fixtureId": "gmod.perf.angle-zero.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "partial", "detail": "Matched a partial-answer pattern.", "finalResponse": "```lua\n-- Slow: allocates a new Angle every call.\nlocal function slow()\n return Angle()\nend\n\n-- Fast: reuse one scratch Angle and zero it in place.\nlocal scratchAng = Angle()\n\nlocal function fast()\n scratchAng:SetUnpacked(0, 0, 0)\n return scratchAng\nend\n```\nReason: Reusing one `Angle` avoids per-call allocation and GC churn; zeroing the same object in place is cheaper than constructing a new `Angle()` on every hot-path call.", "durationMs": 14208, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f6bc97bf98280d62", "startedAt": "2026-07-10T17:25:41.374Z", "completedAt": "2026-07-10T17:25:55.584Z", "answerBytes": 437, "answerChars": 437, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 110, "totalTokens": 181, "source": "estimated" } }, { "fixtureId": "gmod.perf.config-var.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "incorrect", "detail": "Matched an incorrect-answer pattern.", "finalResponse": "```lua\nlocal addonColor = myaddon.config.color\nlocal color = addonColor\n```\nReason: A local is fastest in Lua; `myaddon.config.color` does multiple table lookups, and a global is also a table lookup through `_G`.", "durationMs": 10014, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "bcc4ff8eb162b699", "startedAt": "2026-07-10T17:25:42.747Z", "completedAt": "2026-07-10T17:25:52.762Z", "answerBytes": 212, "answerChars": 212, "exitCode": 0, "usage": { "promptTokens": 82, "completionTokens": 53, "totalTokens": 135, "source": "estimated" } }, { "fixtureId": "gmod.perf.darkrpvar.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal jobName = ply:getDarkRPVar(\"job\") -- preferred over ply:GetNWString(\"job\")\n```\nReason: `job` is a DarkRPVar, so `getDarkRPVar` is the supported DarkRP API and does not depend on generic NWString state.", "durationMs": 57488, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "02a37daa699cded3", "startedAt": "2026-07-10T17:25:43.705Z", "completedAt": "2026-07-10T17:26:41.194Z", "answerBytes": 214, "answerChars": 214, "exitCode": 0, "usage": { "promptTokens": 66, "completionTokens": 54, "totalTokens": 120, "source": "estimated" } }, { "fixtureId": "gmod.perf.disttosqr.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal maxDist = 256\n\nif vec1:DistToSqr(vec2) <= maxDist * maxDist then\n -- within range\nend\n```\nReason: `DistToSqr` avoids the square root used by actual distance calculation, so it is the preferred and faster method for threshold checks when you compare against the squared threshold.", "durationMs": 8164, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "a5772470c6859cdc", "startedAt": "2026-07-10T17:25:43.946Z", "completedAt": "2026-07-10T17:25:52.111Z", "answerBytes": 295, "answerChars": 295, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 74, "totalTokens": 145, "source": "estimated" } }, { "fixtureId": "gmod.perf.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor _, ent in ents.Iterator() do\n -- Visit every entity here.\nend\n```\nReason: `ents.Iterator()` is the wiki-recommended efficient full-entity iterator because it walks the engine list directly without allocating the temporary table that `ents.GetAll()` creates; use `ents.GetAll()` only if you specifically need a snapshot table.", "durationMs": 29364, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "306d3f1d05f28246", "startedAt": "2026-07-10T17:25:44.052Z", "completedAt": "2026-07-10T17:26:13.417Z", "answerBytes": 339, "answerChars": 339, "exitCode": 0, "usage": { "promptTokens": 66, "completionTokens": 85, "totalTokens": 151, "source": "estimated" } }, { "fixtureId": "gmod.perf.find-ents-near.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain 0 through 0 candidate loops.", "finalResponse": "```lua\nlocal radius = 512\nlocal nearby = ents.FindInSphere(ply:GetPos(), radius)\n\nfor i = #nearby, 1, -1 do\n local ent = nearby[i]\n if not IsValid(ent) or ent == ply then\n nearby[i] = nearby[#nearby]\n nearby[#nearby] = nil\n end\nend\n\n-- `nearby` now contains valid entities near the player.\n```\nReason: `ents.FindInSphere` is one of the commonly faster built-in `FindIn*` choices for a cheap near-player query because it stays engine-side, uses spatial partitioning, and matches a radius lookup directly without `FindInPVS` visibility semantics.", "durationMs": 11723, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "07c583fc3d9fe09f", "startedAt": "2026-07-10T17:25:44.529Z", "completedAt": "2026-07-10T17:25:56.254Z", "answerBytes": 567, "answerChars": 567, "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14914, "completionTokens": 506, "totalTokens": 15420, "reasoningTokens": 332, "cachedTokens": 14336 } }, { "fixtureId": "gmod.perf.find-players-near.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal player_GetAll = player.GetAll\n\nlocal function FindPlayersNear(pos, radius)\n local radiusSqr = radius * radius\n local players = player_GetAll()\n local out, n = {}, 0\n\n for i = 1, #players do\n local ply = players[i]\n if ply:GetPos():DistToSqr(pos) <= radiusSqr then\n n = n + 1\n out[n] = ply\n end\n end\n\n return out\nend\n```\nReason: This scans only players instead of all entities, avoids per-entity class checks, and uses squared distance to skip the expensive square root.", "durationMs": 46343, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "d139ba2e8ee38a2f", "startedAt": "2026-07-10T17:25:44.716Z", "completedAt": "2026-07-10T17:26:31.061Z", "answerBytes": 543, "answerChars": 543, "exitCode": 0, "usage": { "promptTokens": 79, "completionTokens": 136, "totalTokens": 215, "source": "estimated" } }, { "fixtureId": "gmod.perf.for-vs-while.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor i = 1, n do\n -- work\nend\n```\nReason: Numeric `for` is generally faster for 1..n counting in Garry's Mod Lua because it uses the language's built-in counted-loop path and avoids the extra manual condition and increment work of a `while` loop.", "durationMs": 9279, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "b3401a190a8aecf0", "startedAt": "2026-07-10T17:25:44.981Z", "completedAt": "2026-07-10T17:25:54.261Z", "answerBytes": 255, "answerChars": 255, "exitCode": 0, "usage": { "promptTokens": 65, "completionTokens": 64, "totalTokens": 129, "source": "estimated" } }, { "fixtureId": "gmod.perf.hook-once.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\n-- Faster: one hook dispatch per tick, then iterate players inside the listener.\nhook.Add(\"Tick\", \"NotifyAllPlayersOnce\", function()\n hook.Run(\"MyPerTickPlayers\", player.GetAll())\nend)\n\nhook.Add(\"MyPerTickPlayers\", \"HandlePlayers\", function(players)\n for i = 1, #players do\n local ply = players[i]\n -- custom per-player logic here\n end\nend)\n\n-- Slower:\n-- hook.Add(\"Tick\", \"NotifyEachPlayerSeparately\", function()\n-- local players = player.GetAll()\n-- for i = 1, #players do\n-- hook.Run(\"MyPerTickPlayer\", players[i])\n-- end\n-- end)\n```\nReason: `hook.Run` has dispatch overhead, so calling it once per tick and looping players inside the handler does less total work than re-entering the hook system once per player.", "durationMs": 14651, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "ed1d3e09249fc8bb", "startedAt": "2026-07-10T17:25:45.160Z", "completedAt": "2026-07-10T17:25:59.813Z", "answerBytes": 766, "answerChars": 766, "exitCode": 0, "usage": { "promptTokens": 84, "completionTokens": 192, "totalTokens": 276, "source": "estimated" } }, { "fixtureId": "gmod.perf.hudpaint-cache.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nhook.Add(\"HUDPaint\", \"DrawHealth\", function()\n local ply = LocalPlayer()\n if not IsValid(ply) then return end\n\n local health = ply:Health()\n local healthText = tostring(health)\n\n draw.RoundedBox(4, 20, 20, health * 2, 24, Color(200, 60, 60))\n draw.SimpleText(healthText, \"Trebuchet24\", 28, 22, color_white)\n draw.SimpleTextOutlined(healthText, \"Trebuchet24\", 28, 22, color_white, TEXT_ALIGN_LEFT, TEXT_ALIGN_TOP, 1, color_black)\nend)\n```\nReason: Cache `LocalPlayer():Health()` once per `HUDPaint` so the three draw calls reuse one value instead of repeating the method lookup and call in the same frame.", "durationMs": 15524, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3def450f3158ceff", "startedAt": "2026-07-10T17:25:45.748Z", "completedAt": "2026-07-10T17:26:01.274Z", "answerBytes": 631, "answerChars": 631, "exitCode": 0, "usage": { "promptTokens": 76, "completionTokens": 158, "totalTokens": 234, "source": "estimated" } }, { "fixtureId": "gmod.perf.local-color.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal WHITE = Color(255, 255, 255)\n\nfunction PANEL:Paint(w, h)\n for i = 1, 100 do\n surface.SetDrawColor(WHITE)\n surface.DrawRect(0, i * 4, w, 2)\n end\nend\n```\nReason: `Color(255, 255, 255)` creates a new color object each iteration; reusing one cached color removes per-frame allocations and GC churn.", "durationMs": 16287, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "16a7bed1b931d467", "startedAt": "2026-07-10T17:25:45.967Z", "completedAt": "2026-07-10T17:26:02.255Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14890, "completionTokens": 358, "totalTokens": 15248, "reasoningTokens": 235, "cachedTokens": 14336 }, "answerBytes": 327, "answerChars": 327 }, { "fixtureId": "gmod.perf.local-vs-global.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal sin = math.sin\n\nfor i = 1, 1000000 do\n local y = sin(i)\n -- hot-loop work\nend\n```\nReason: Caching `math.sin` in a local avoids repeated table/global lookups inside the hot loop, which is a small but common Lua micro-optimization.", "durationMs": 13584, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "c116d8c651190647", "startedAt": "2026-07-10T17:25:46.159Z", "completedAt": "2026-07-10T17:25:59.745Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14894, "completionTokens": 217, "totalTokens": 15111, "reasoningTokens": 127, "cachedTokens": 14336 }, "answerBytes": 248, "answerChars": 248 }, { "fixtureId": "gmod.perf.localplayer-cache.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "partial", "detail": "Matched a partial-answer pattern.", "finalResponse": "```lua\nlocal getLocalPlayer = LocalPlayer\nlocal cachedLocalPlayer\n\nlocal function LP()\n local ply = cachedLocalPlayer\n if not IsValid(ply) then\n ply = getLocalPlayer()\n if not IsValid(ply) then return nil end\n cachedLocalPlayer = ply\n end\n return ply\nend\n\nhook.Add(\"Think\", \"HotPathExample\", function()\n local ply = LP()\n if not ply then return end\n\n if ply:Alive() then\n -- use ply here instead of calling LocalPlayer() again\n end\nend)\n```\nReason: It avoids repeated global/function lookup and player resolution on hot paths by caching the local player and only refreshing when that entity becomes invalid.", "durationMs": 22563, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "7398105af47b079e", "startedAt": "2026-07-10T17:25:46.754Z", "completedAt": "2026-07-10T17:26:09.319Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14897, "completionTokens": 684, "totalTokens": 15581, "reasoningTokens": 503, "cachedTokens": 14336 }, "answerBytes": 659, "answerChars": 659 }, { "fixtureId": "gmod.perf.looking-at.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "partial", "detail": "Matched a partial-answer pattern.", "finalResponse": "```lua\nlocal function RoughlyLookingAtEachOther(a, b, dotThreshold)\n dotThreshold = dotThreshold or 0.8 -- higher = stricter\n\n local aToB = (b:EyePos() - a:EyePos()):GetNormalized()\n local bToA = -aToB\n\n return a:GetAimVector():Dot(aToB) >= dotThreshold\n and b:GetAimVector():Dot(bToA) >= dotThreshold\nend\n```\nReason: Dot products avoid `math.acos`, so this stays fast and directly checks whether each player's aim vector points close enough toward the other player's eye position.", "durationMs": 21925, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "20b22f678a4b08d7", "startedAt": "2026-07-10T17:25:47.248Z", "completedAt": "2026-07-10T17:26:09.175Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14909, "completionTokens": 673, "totalTokens": 15582, "reasoningTokens": 516, "cachedTokens": 14336 }, "answerBytes": 499, "answerChars": 499 }, { "fixtureId": "gmod.perf.meta-vs-arg.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nSomeHelper(ply)\n```\nReason: `SomeHelper(ply)` is generally faster for a helper you control because it can be a direct function call and avoids the extra table/metatable method lookup behind `ply:SomeHelper()`, though the gap is usually small outside hot paths.", "durationMs": 17635, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "de332f7356830dfa", "startedAt": "2026-07-10T17:25:47.428Z", "completedAt": "2026-07-10T17:26:05.064Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14901, "completionTokens": 411, "totalTokens": 15312, "reasoningTokens": 332, "cachedTokens": 0 }, "answerBytes": 267, "answerChars": 267 }, { "fixtureId": "gmod.perf.msg-vs-print.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nMsgN(\"[debug] hot path hit\")\n```\nReason: `MsgN` is faster than `print` in Garry's Mod because it writes directly to the console with less argument-handling and formatting overhead.", "durationMs": 23345, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "6cbb4e4b2f19dd24", "startedAt": "2026-07-10T17:25:47.640Z", "completedAt": "2026-07-10T17:26:10.986Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14894, "completionTokens": 583, "totalTokens": 15477, "reasoningTokens": 516, "cachedTokens": 14336 }, "answerBytes": 187, "answerChars": 187 }, { "fixtureId": "gmod.perf.pairs-ipairs-for.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal nums = someSequentialNumberTable\n\nfor i = 1, #nums do\n local v = nums[i]\n -- use v\nend\n```\nReason: `numeric for` avoids iterator overhead and direct integer indexing is typically faster than `ipairs`, while `pairs` is for generic table traversal and is usually the slowest for a dense sequential array.", "durationMs": 16629, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "de5e4875fcf3e600", "startedAt": "2026-07-10T17:25:50.765Z", "completedAt": "2026-07-10T17:26:07.395Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14902, "completionTokens": 375, "totalTokens": 15277, "reasoningTokens": 277, "cachedTokens": 14336 }, "answerBytes": 321, "answerChars": 321 }, { "fixtureId": "gmod.perf.plain-lua-vs-fn.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal IsValid = IsValid\n\nlocal function getPoliceNames(players)\n local out = {}\n local n = 0\n\n for i = 1, #players do\n local ply = players[i]\n\n if IsValid(ply) and ply:Alive() and ply:isCP() then\n n = n + 1\n out[n] = ply:Nick()\n end\n end\n\n return out\nend\n\nlocal policeNames = getPoliceNames(player.GetAll())\n```\nReason: A single indexed loop fuses filter and map into one pass, avoids fn closure/combinator overhead and intermediate tables, and is usually faster in GMod Lua.", "durationMs": 44243, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "ead0a98b3ff6a388", "startedAt": "2026-07-10T17:25:52.112Z", "completedAt": "2026-07-10T17:26:36.357Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14916, "completionTokens": 1854, "totalTokens": 16770, "reasoningTokens": 1683, "cachedTokens": 14336 }, "answerBytes": 540, "answerChars": 540 }, { "fixtureId": "gmod.perf.player-cache-hooks.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal cachedPlayers = player.GetAll()\n\nhook.Add(\"PlayerInitialSpawn\", \"MyPlayerCacheAdd\", function(ply)\n cachedPlayers[#cachedPlayers + 1] = ply\nend)\n\nhook.Add(\"PlayerDisconnected\", \"MyPlayerCacheRemove\", function(ply)\n for i = 1, #cachedPlayers do\n if cachedPlayers[i] == ply then\n table.remove(cachedPlayers, i)\n break\n end\n end\nend)\n\nhook.Add(\"Tick\", \"MyHotPlayerLoop\", function()\n for i = 1, #cachedPlayers do\n local ply = cachedPlayers[i]\n\n -- per-player work here\n -- Example: if ply:Alive() then ... end\n end\nend)\n```\nReason: `player.GetAll()` creates/traverses a player list every call, so moving that work to join/leave hooks keeps the hot `Tick` loop to a cheap cached-table scan.", "durationMs": 69771, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "a834817afe9e6c25", "startedAt": "2026-07-10T17:25:52.411Z", "completedAt": "2026-07-10T17:27:02.183Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14911, "completionTokens": 3300, "totalTokens": 18211, "reasoningTokens": 3069, "cachedTokens": 14336 }, "answerBytes": 769, "answerChars": 769 }, { "fixtureId": "gmod.perf.ply-table-index.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "incorrect", "detail": "Did not match a fixture answer pattern.", "finalResponse": "```lua\nlocal tempByPlayerId = {}\n\nlocal function SetTemp(ply, value)\n tempByPlayerId[ply:EntIndex()] = value\nend\n\nlocal function GetTemp(ply)\n return tempByPlayerId[ply:EntIndex()]\nend\n\nhook.Add(\"PlayerDisconnected\", \"ClearTempByPlayerId\", function(ply)\n tempByPlayerId[ply:EntIndex()] = nil\nend)\n\n-- hot read path\nlocal id = ply:EntIndex()\nlocal value = tempByPlayerId[id]\n```\nReason: A plain Lua table indexed by `EntIndex()` avoids Garry's Mod entity field/metatable lookup on each read, so hot temporary per-player reads are faster than `ply.someField`.", "durationMs": 40869, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f18cc80df8999147", "startedAt": "2026-07-10T17:25:52.763Z", "completedAt": "2026-07-10T17:26:33.633Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14896, "completionTokens": 1708, "totalTokens": 16604, "reasoningTokens": 1524, "cachedTokens": 14336 }, "answerBytes": 566, "answerChars": 566 }, { "fixtureId": "gmod.perf.seq-tables.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal values = {\n \"alpha\",\n \"beta\",\n \"gamma\",\n \"delta\",\n}\n\nfor i = 1, #values do\n local value = values[i]\n -- use value\nend\n```\nReason: Sequential number-indexed arrays use Lua's array part and a numeric `for` loop, which iterates faster than traversing string-keyed maps with `pairs`. ", "durationMs": 19368, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "4778b0220c3f91a3", "startedAt": "2026-07-10T17:25:53.746Z", "completedAt": "2026-07-10T17:26:13.115Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14899, "completionTokens": 529, "totalTokens": 15428, "reasoningTokens": 419, "cachedTokens": 14336 }, "answerBytes": 311, "answerChars": 311 }, { "fixtureId": "gmod.perf.setdrawcolor-split.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal r, g, b, a = 255, 120, 40, 255\n\nhook.Add(\"HUDPaint\", \"HotDraw\", function()\n surface.SetDrawColor(r, g, b, a)\n surface.DrawRect(100, 100, 200, 40)\nend)\n```\nReason: Passing raw channel numbers avoids the extra Color object/table path and any per-frame allocation, so it is the lower-overhead call style for hot draw code.", "durationMs": 19028, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "d110fde8a37bbe59", "startedAt": "2026-07-10T17:25:54.262Z", "completedAt": "2026-07-10T17:26:13.291Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14902, "completionTokens": 497, "totalTokens": 15399, "reasoningTokens": 373, "cachedTokens": 14336 }, "answerBytes": 338, "answerChars": 338 }, { "fixtureId": "gmod.perf.sqlite-vs-mysql.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal steamId64 = \"76561198000000000\"\nlocal row = sql.QueryRow(\"SELECT cash FROM players WHERE steamid64 = \" .. sql.SQLStr(steamId64))\n```\nReason: Local SQLite via GMod's `sql` library is typically lower latency on a single server because it avoids network round trips and remote database overhead.", "durationMs": 16725, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "4a3f5d2a20d71ac9", "startedAt": "2026-07-10T17:25:55.585Z", "completedAt": "2026-07-10T17:26:12.312Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14920, "completionTokens": 379, "totalTokens": 15299, "reasoningTokens": 284, "cachedTokens": 14336 }, "answerBytes": 305, "answerChars": 305 }, { "fixtureId": "gmod.perf.steamid-cache.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal PLAYER = FindMetaTable(\"Player\")\n\nfunction PLAYER:GetSteamIDCached()\n local sid = self._cachedSteamID\n if sid == nil then\n sid = self:SteamID()\n self._cachedSteamID = sid\n end\n return sid\nend\n\nhook.Add(\"PlayerInitialSpawn\", \"WarmSteamIDCache\", function(ply)\n ply._cachedSteamID = ply:SteamID()\nend)\n\nhook.Add(\"PlayerDisconnected\", \"ClearSteamIDCache\", function(ply)\n ply._cachedSteamID = nil\nend)\n\n-- Hot path\nlocal function ProcessPlayers(players)\n for i = 1, #players do\n local ply = players[i]\n local sid = ply._cachedSteamID or ply:GetSteamIDCached()\n\n -- use sid here\n end\nend\n```\nReason: It pays the `SteamID()` method call once per player, then reuses a plain field read in the hot path, which is cheaper than repeated userdata method dispatch.", "durationMs": 22650, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "e8a94aa454154982", "startedAt": "2026-07-10T17:25:56.255Z", "completedAt": "2026-07-10T17:26:18.906Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14893, "completionTokens": 763, "totalTokens": 15656, "reasoningTokens": 516, "cachedTokens": 14336 }, "answerBytes": 823, "answerChars": 823 }, { "fixtureId": "gmod.perf.string-table-concat.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain 0 through 0 candidate loops.", "finalResponse": "```lua\n-- Slow:\n-- local big = \"\"\n-- for i = 1, 10000 do\n-- big = big .. (\"piece_\" .. i)\n-- end\n\n-- Fast:\nlocal parts = {}\n\nfor i = 1, 10000 do\n parts[i] = \"piece_\" .. i\nend\n\nlocal big = table.concat(parts, \"\")\n```\nReason: `table.concat` builds the final string in one pass, avoiding the repeated reallocation and full-string copying caused by `..` inside a loop.", "durationMs": 20748, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "b0cf9e04ae7c5dd9", "startedAt": "2026-07-10T17:25:59.746Z", "completedAt": "2026-07-10T17:26:20.494Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14894, "completionTokens": 524, "totalTokens": 15418, "reasoningTokens": 381, "cachedTokens": 14336 }, "answerBytes": 370, "answerChars": 370 }, { "fixtureId": "gmod.perf.surface-drawrect.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nsurface.SetDrawColor(r, g, b, a)\nsurface.DrawRect(x, y, w, h)\n```\nReason: `surface.DrawRect` is a more direct filled-rectangle call and avoids the extra helper/corner-handling overhead in `draw.RoundedBox(0, ...)`. ", "durationMs": 16607, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "696e8de769d9ad0e", "startedAt": "2026-07-10T17:25:59.814Z", "completedAt": "2026-07-10T17:26:16.422Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14900, "completionTokens": 405, "totalTokens": 15305, "reasoningTokens": 322, "cachedTokens": 14336 }, "answerBytes": 222, "answerChars": 222 }, { "fixtureId": "gmod.perf.surface-text.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal SetFont = surface.SetFont\nlocal SetTextColor = surface.SetTextColor\nlocal SetTextPos = surface.SetTextPos\nlocal DrawText = surface.DrawText\n\nlocal FONT = \"DermaDefaultBold\"\nlocal TEXT = \"Single line every frame\"\nlocal X, Y = 24, 24\n\nhook.Add(\"HUDPaint\", \"FastSingleLineText\", function()\n SetFont(FONT)\n SetTextColor(255, 255, 255, 255)\n SetTextPos(X, Y)\n DrawText(TEXT)\nend)\n```\nReason: Direct `surface` text calls are faster for a single-line draw because they avoid the extra wrapper work in `draw.DrawText` and the multiple text passes used by `draw.SimpleTextOutlined`. ", "durationMs": 31623, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "95d9fe4918ee4197", "startedAt": "2026-07-10T17:26:00.017Z", "completedAt": "2026-07-10T17:26:31.641Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14895, "completionTokens": 1223, "totalTokens": 16118, "reasoningTokens": 1034, "cachedTokens": 14336 }, "answerBytes": 599, "answerChars": 599 }, { "fixtureId": "gmod.perf.table-count-hash.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal len = #myTable\n```\nReason: `#` is the preferred way to get the length of a sequential array-like table; `table.Count` is for non-sequential keys and does extra work.", "durationMs": 12617, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3e9bbfc8452e05e3", "startedAt": "2026-07-10T17:26:01.275Z", "completedAt": "2026-07-10T17:26:13.894Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14889, "completionTokens": 175, "totalTokens": 15064, "reasoningTokens": 111, "cachedTokens": 14336 }, "answerBytes": 178, "answerChars": 178 }, { "fixtureId": "gmod.perf.table-empty.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nt = {}\n```\nReason: `t = {}` is generally faster because it replaces the large sequential table instead of iterating and niling every element like `table.Empty(t)`; only keep `table.Empty(t)` when other code must retain the same table reference.", "durationMs": 17776, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3a6344977f5bd475", "startedAt": "2026-07-10T17:26:02.256Z", "completedAt": "2026-07-10T17:26:20.033Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14898, "completionTokens": 420, "totalTokens": 15318, "reasoningTokens": 344, "cachedTokens": 14336 }, "answerBytes": 251, "answerChars": 251 }, { "fixtureId": "gmod.perf.table-hasvalue.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal allowed = {\n pistol = true,\n shotgun = true,\n crowbar = true,\n}\n\nif allowed[weaponClass] then\n -- value exists\nend\n```\nReason: A set-style table uses direct key lookup with average O(1) cost, while `table.HasValue` walks a list linearly with O(n) cost.", "durationMs": 20739, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "7d064428a873fdfc", "startedAt": "2026-07-10T17:26:05.065Z", "completedAt": "2026-07-10T17:26:25.805Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14898, "completionTokens": 461, "totalTokens": 15359, "reasoningTokens": 362, "cachedTokens": 14336 }, "answerBytes": 277, "answerChars": 277 }, { "fixtureId": "gmod.perf.table-literal.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\n-- Faster for a small table with known fields:\nlocal t = {\n x = 10,\n y = 20,\n name = \"npc\",\n}\n\n-- Slower:\n-- local t = {}\n-- t.x = 10\n-- t.y = 20\n-- t.name = \"npc\"\n```\nReason: Putting known fields directly in the table constructor is faster because Lua can build the table in one step with fewer bytecode operations and fewer hash inserts than creating `{}` and assigning each field afterward.", "durationMs": 21596, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f95a8fc610d720ad", "startedAt": "2026-07-10T17:26:07.396Z", "completedAt": "2026-07-10T17:26:28.993Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14897, "completionTokens": 651, "totalTokens": 15548, "reasoningTokens": 516, "cachedTokens": 14336 }, "answerBytes": 409, "answerChars": 409 }, { "fixtureId": "gmod.perf.table-random.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "incorrect", "detail": "Did not match a fixture answer pattern.", "finalResponse": "```lua\nlocal function randomSequential(t)\n local n = #t\n if n == 0 then return nil end\n return t[math.random(n)]\nend\n\nlocal value = randomSequential(myTable)\n```\nReason: `table.Random` handles general tables, but a sequential table can skip that overhead and pick a random numeric index directly with `t[math.random(#t)]`. ", "durationMs": 17175, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f3e090701953d4a9", "startedAt": "2026-07-10T17:26:09.175Z", "completedAt": "2026-07-10T17:26:26.352Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14891, "completionTokens": 349, "totalTokens": 15240, "reasoningTokens": 245, "cachedTokens": 14336 }, "answerBytes": 332, "answerChars": 332 }, { "fixtureId": "gmod.perf.x-times-x.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal y = x * x\n```\nReason: `x * x` is the fast preferred form in a tight loop because it avoids the slower exponentiation operator `x ^ 2`. ", "durationMs": 13176, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "79d48008f6b5174f", "startedAt": "2026-07-10T17:26:09.320Z", "completedAt": "2026-07-10T17:26:22.497Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14888, "completionTokens": 209, "totalTokens": 15097, "reasoningTokens": 150, "cachedTokens": 14336 }, "answerBytes": 148, "answerChars": 148 }, { "fixtureId": "gmod.player-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Uses player.Iterator() (wiki-recommended Lua-cached iterator).", "finalResponse": "```lua\nfor _, ply in player.Iterator() do\n -- ...\nend\n```\nReason: `player.Iterator()` visits connected players directly without creating the temporary table that `player.GetAll()` returns.", "durationMs": 21353, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 2, "rubricVersion": "4", "promptHash": "fe3d7469535a0ca2", "startedAt": "2026-07-10T17:26:10.987Z", "completedAt": "2026-07-10T17:26:32.341Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14904, "completionTokens": 577, "totalTokens": 15481, "reasoningTokens": 516, "cachedTokens": 14336 }, "answerBytes": 191, "answerChars": 191 }, { "fixtureId": "gmod.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor ent in ents.Iterator() do\n if ent:GetClass() == \"prop_physics\" then\n -- handle ent here\n end\nend\n```\nReason: `ents.Iterator()` walks entities directly, so it avoids building a full entity table first.", "durationMs": 12457, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "3990c92253db85bb", "startedAt": "2026-07-10T17:26:12.313Z", "completedAt": "2026-07-10T17:26:24.770Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14543, "completionTokens": 220, "totalTokens": 14763, "reasoningTokens": 145, "cachedTokens": 0 }, "answerBytes": 224, "answerChars": 224 }, { "fixtureId": "gmod.hook-add.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Registers a named hook with hook.Add.", "finalResponse": "```lua\nhook.Add(\"Think\", \"MyAddon.Think\", function()\nend)\n```\nReason: Registers a Think hook callback under the `MyAddon.Think` identifier so Garry's Mod can invoke it every frame.", "durationMs": 11293, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "12d2222fdb7d6b90", "startedAt": "2026-07-10T17:26:13.116Z", "completedAt": "2026-07-10T17:26:24.410Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14549, "completionTokens": 74, "totalTokens": 14623, "reasoningTokens": 8, "cachedTokens": 0 }, "answerBytes": 180, "answerChars": 180 }, { "fixtureId": "gmod.isvalid.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Uses IsValid for entity/object validity.", "finalResponse": "```lua\nif IsValid(ent) then\n -- safe to call methods on ent here\nend\n```\nReason: `IsValid` checks that the entity reference still points to a live, usable entity before you call methods on it.", "durationMs": 10680, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "35527312e59dfc13", "startedAt": "2026-07-10T17:26:13.292Z", "completedAt": "2026-07-10T17:26:23.973Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14544, "completionTokens": 75, "totalTokens": 14619, "reasoningTokens": 8, "cachedTokens": 0 }, "answerBytes": 195, "answerChars": 195 }, { "fixtureId": "gmod.net-receivers.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nnet.Receive(\"MyAddon.Sync\", function(len, ply)\n -- handle the incoming message here\nend)\n```\nReason: `net.Receive` registers the client or server callback for an incoming Garry's Mod net message by name.", "durationMs": 15321, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "24d54ee8366753a6", "startedAt": "2026-07-10T17:26:13.418Z", "completedAt": "2026-07-10T17:26:28.741Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14539, "completionTokens": 80, "totalTokens": 14619, "reasoningTokens": 10, "cachedTokens": 0 }, "answerBytes": 213, "answerChars": 213 }, { "fixtureId": "gmod.perf.angle-zero.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "incorrect", "detail": "Did not match a fixture answer pattern.", "finalResponse": "```lua\nlocal zero = angle_zero\n\n-- hot path\nlocal a = zero\n-- use `a` as a zeroed Angle without allocating a new one\n\n-- slower\nlocal b = Angle(0, 0, 0)\n``` \nReason: `angle_zero` is a shared zeroed Angle, so reusing it avoids repeated `Angle()` allocations in hot paths.", "durationMs": 13781, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f6bc97bf98280d62", "startedAt": "2026-07-10T17:26:13.895Z", "completedAt": "2026-07-10T17:26:27.677Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14543, "completionTokens": 189, "totalTokens": 14732, "reasoningTokens": 90, "cachedTokens": 0 }, "answerBytes": 270, "answerChars": 270 }, { "fixtureId": "gmod.perf.config-var.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal myaddon_config_color = Color(255, 255, 255)\n\nlocal color = myaddon_config_color\n```\nReason: a flat local avoids repeated table lookups, so it is faster for a hot, frequently read setting than `myaddon.config.color`.", "durationMs": 10712, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "bcc4ff8eb162b699", "startedAt": "2026-07-10T17:26:16.423Z", "completedAt": "2026-07-10T17:26:27.136Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14546, "completionTokens": 101, "totalTokens": 14647, "reasoningTokens": 23, "cachedTokens": 0 }, "answerBytes": 228, "answerChars": 228 }, { "fixtureId": "gmod.perf.darkrpvar.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nply:getDarkRPVar(\"job\")\n```\nReason: `getDarkRPVar(\"job\")` is DarkRP’s job API and is the preferred way to read a player’s job name; `GetNWString(\"job\")` is the lower-level networked-value fallback and may not match DarkRP’s own state handling.", "durationMs": 5862, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "02a37daa699cded3", "startedAt": "2026-07-10T17:26:18.907Z", "completedAt": "2026-07-10T17:26:24.770Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14544, "completionTokens": 103, "totalTokens": 14647, "reasoningTokens": 14, "cachedTokens": 0 }, "answerBytes": 256, "answerChars": 250 }, { "fixtureId": "gmod.perf.disttosqr.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nif vec1:DistToSqr(vec2) < threshold * threshold then\n -- within range\nend\n```\nReason: `DistToSqr` avoids the square root, so it is the preferred method for a threshold check.", "durationMs": 4265, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "a5772470c6859cdc", "startedAt": "2026-07-10T17:26:20.034Z", "completedAt": "2026-07-10T17:26:24.300Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14545, "completionTokens": 78, "totalTokens": 14623, "reasoningTokens": 8, "cachedTokens": 0 }, "answerBytes": 184, "answerChars": 184 }, { "fixtureId": "gmod.perf.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor ent in ents.Iterator() do\n -- visit every entity here\nend\n```\nReason: `ents.Iterator()` is the wiki-recommended efficient way to traverse all entities when you want to iterate them directly without building a full `ents.GetAll()` table first.", "durationMs": 4992, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "306d3f1d05f28246", "startedAt": "2026-07-10T17:26:20.495Z", "completedAt": "2026-07-10T17:26:25.489Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14541, "completionTokens": 143, "totalTokens": 14684, "reasoningTokens": 68, "cachedTokens": 13824 }, "answerBytes": 256, "answerChars": 256 }, { "fixtureId": "gmod.perf.find-ents-near.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain 0 through 0 candidate loops.", "finalResponse": "```lua\nlocal ply = Entity(1) -- or any player entity\nif not IsValid(ply) then return end\n\nlocal pos = ply:GetPos()\nlocal radius = 512\nlocal nearby = ents.FindInSphere(pos, radius)\n\nfor i = 1, #nearby do\n local ent = nearby[i]\n if IsValid(ent) and ent ~= ply then\n -- cheap per-entity filtering here\n end\nend\n```\nReason: `ents.FindInSphere` is one of the faster built-in spatial queries in common GMod benchmarks, and it is the simplest cheap broad-phase query for entities near a player.", "durationMs": 5493, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "07c583fc3d9fe09f", "startedAt": "2026-07-10T17:26:22.498Z", "completedAt": "2026-07-10T17:26:27.993Z", "answerBytes": 503, "answerChars": 503, "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14564, "completionTokens": 204, "totalTokens": 14768, "reasoningTokens": 47, "cachedTokens": 0 } }, { "fixtureId": "gmod.perf.find-players-near.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal function PlayersInRadius(pos, radius)\n local radiusSqr = radius * radius\n local out = {}\n\n for _, ply in ipairs(player.GetAll()) do\n if pos:DistToSqr(ply:GetPos()) <= radiusSqr then\n out[#out + 1] = ply\n end\n end\n\n return out\nend\n``` \nReason: `player.GetAll()` gives only players, so it avoids building and filtering a mixed entity list from `ents.FindInSphere`; `DistToSqr` also skips the square-root cost.", "durationMs": 5409, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "d139ba2e8ee38a2f", "startedAt": "2026-07-10T17:26:23.973Z", "completedAt": "2026-07-10T17:26:29.384Z", "answerBytes": 460, "answerChars": 460, "exitCode": 0, "usage": { "promptTokens": 79, "completionTokens": 115, "totalTokens": 194, "source": "estimated" } }, { "fixtureId": "gmod.perf.for-vs-while.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor i = 1, n do\n -- tight counting loop\nend\n```\nReason: Numeric `for` is generally faster than `while` in Garry's Mod Lua because the VM optimizes it as a dedicated loop form with less per-iteration overhead.", "durationMs": 4115, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "b3401a190a8aecf0", "startedAt": "2026-07-10T17:26:24.301Z", "completedAt": "2026-07-10T17:26:28.418Z", "answerBytes": 218, "answerChars": 218, "exitCode": 0, "usage": { "promptTokens": 65, "completionTokens": 55, "totalTokens": 120, "source": "estimated" } }, { "fixtureId": "gmod.perf.hook-once.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\n-- Faster: one hook call per tick, pass the whole player list once.\nlocal players = player.GetAll()\nhook.Run(\"MyCustomTick\", players)\n\n-- Slower: one hook call per player per tick.\nfor i = 1, #players do\n hook.Run(\"MyCustomTick\", players[i])\nend\n``` \nReason: One `hook.Run` call avoids repeating hook dispatch overhead for every player, so it is cheaper than calling it once per player each tick.", "durationMs": 10074, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "ed1d3e09249fc8bb", "startedAt": "2026-07-10T17:26:24.412Z", "completedAt": "2026-07-10T17:26:34.488Z", "answerBytes": 406, "answerChars": 406, "exitCode": 0, "usage": { "promptTokens": 84, "completionTokens": 102, "totalTokens": 186, "source": "estimated" } }, { "fixtureId": "gmod.perf.hudpaint-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal lp = LocalPlayer()\nlocal hp = lp:Health()\n\n-- use hp for all three draw calls below\n-- draw first bar based on hp\n-- draw second bar based on hp\n-- draw third bar based on hp\n```\nReason: Cache `LocalPlayer()` and `:Health()` once per frame so the hook does one player lookup and one health read instead of repeating both three times.", "durationMs": 3994, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3def450f3158ceff", "startedAt": "2026-07-10T17:26:24.771Z", "completedAt": "2026-07-10T17:26:28.767Z", "answerBytes": 346, "answerChars": 346, "exitCode": 0, "usage": { "promptTokens": 76, "completionTokens": 87, "totalTokens": 163, "source": "estimated" } }, { "fixtureId": "gmod.perf.local-color.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal white = color_white\n\nfor i = 1, count do\n surface.SetDrawColor(white)\n -- paint work here\nend\n```\nReason: Reuses Garry's Mod's shared `color_white` instead of allocating a new `Color(255, 255, 255)` table every iteration.", "durationMs": 4261, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "16a7bed1b931d467", "startedAt": "2026-07-10T17:26:24.771Z", "completedAt": "2026-07-10T17:26:29.039Z", "answerBytes": 240, "answerChars": 240, "exitCode": 0, "usage": { "promptTokens": 63, "completionTokens": 60, "totalTokens": 123, "source": "estimated" } }, { "fixtureId": "gmod.perf.local-vs-global.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal sin = math.sin\n\nfor i = 1, 1000000 do\n local y = sin(i)\n -- use y\nend\n``` \nReason: Caching `math.sin` in a local avoids repeated global/table lookup inside the hot loop.", "durationMs": 3744, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "c116d8c651190647", "startedAt": "2026-07-10T17:26:25.490Z", "completedAt": "2026-07-10T17:26:29.235Z", "answerBytes": 188, "answerChars": 188, "exitCode": 0, "usage": { "promptTokens": 62, "completionTokens": 47, "totalTokens": 109, "source": "estimated" } }, { "fixtureId": "gmod.perf.localplayer-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal LocalPlayer = LocalPlayer\n\nhook.Add(\"Think\", \"ExampleCachedLocalPlayer\", function()\n local lp = LocalPlayer()\n if not IsValid(lp) then return end\n\n -- use lp many times here instead of calling LocalPlayer() repeatedly\nend)\n``` \nReason: Caching the global function in a local avoids repeated global table lookup, and storing the player in a local inside the hot path avoids calling `LocalPlayer()` multiple times in the same frame.", "durationMs": 3946, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "7398105af47b079e", "startedAt": "2026-07-10T17:26:25.806Z", "completedAt": "2026-07-10T17:26:29.752Z", "answerBytes": 452, "answerChars": 452, "exitCode": 0, "usage": { "promptTokens": 70, "completionTokens": 113, "totalTokens": 183, "source": "estimated" } }, { "fixtureId": "gmod.perf.looking-at.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "partial", "detail": "Matched a partial-answer pattern.", "finalResponse": "```lua\nlocal function RoughlyLookingAtEachOther(p1, p2, threshold)\n threshold = threshold or 0.90 -- closer to 1.0 = stricter\n\n local a1 = p1:EyeAngles():Forward()\n local a2 = p2:EyeAngles():Forward()\n local to2 = (p2:EyePos() - p1:EyePos()):GetNormalized()\n local to1 = -to2\n\n return a1:Dot(to2) >= threshold and a2:Dot(to1) >= threshold\nend\n```\nReason: Dot products are cheaper than acos and directly test how aligned each player's aim vector is with the direction to the other player.", "durationMs": 4868, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "20b22f678a4b08d7", "startedAt": "2026-07-10T17:26:26.352Z", "completedAt": "2026-07-10T17:26:31.222Z", "answerBytes": 505, "answerChars": 505, "exitCode": 0, "usage": { "promptTokens": 91, "completionTokens": 127, "totalTokens": 218, "source": "estimated" } }, { "fixtureId": "gmod.perf.meta-vs-arg.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nSomeHelper(ply)\n```\nReason: For a helper you control, the free-function style is generally a bit faster and easier to optimize because it avoids method-call lookup and implicit self dispatch; the difference is usually small, but `SomeHelper(ply)` is the preferred style.", "durationMs": 3452, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "de332f7356830dfa", "startedAt": "2026-07-10T17:26:27.137Z", "completedAt": "2026-07-10T17:26:30.590Z", "answerBytes": 277, "answerChars": 277, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 70, "totalTokens": 141, "source": "estimated" } }, { "fixtureId": "gmod.perf.msg-vs-print.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nMsgN(\"debug: hot path hit\")\n```\nReason: `MsgN` writes straight to the console and avoids `print`'s extra formatting overhead, so it is the lighter choice for hot-path debug output.", "durationMs": 4118, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "6cbb4e4b2f19dd24", "startedAt": "2026-07-10T17:26:27.678Z", "completedAt": "2026-07-10T17:26:31.797Z", "answerBytes": 187, "answerChars": 187, "exitCode": 0, "usage": { "promptTokens": 63, "completionTokens": 47, "totalTokens": 110, "source": "estimated" } }, { "fixtureId": "gmod.perf.pairs-ipairs-for.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\n-- Fastest typical loop for a dense array-like table\nfor i = 1, #t do\n local v = t[i]\n -- use v\nend\n```\nReason: Numeric `for` is typically fastest for large sequential tables because it avoids the iterator overhead of `pairs` and `ipairs`.", "durationMs": 4803, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "de5e4875fcf3e600", "startedAt": "2026-07-10T17:26:27.994Z", "completedAt": "2026-07-10T17:26:32.798Z", "answerBytes": 252, "answerChars": 252, "exitCode": 0, "usage": { "promptTokens": 73, "completionTokens": 63, "totalTokens": 136, "source": "estimated" } }, { "fixtureId": "gmod.perf.plain-lua-vs-fn.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\n-- Faster than chaining DarkRP fn combinators: one pass, no temp closures, no temp lists.\nlocal function filterMap(list, keep, map)\n local out = {}\n local n = 0\n\n for i = 1, #list do\n local v = list[i]\n if keep(v, i, list) then\n n = n + 1\n out[n] = map(v, i, list)\n end\n end\n\n return out\nend\n\n-- Example:\n-- local result = filterMap(players.GetAll(),\n-- function(p) return p:Alive() end,\n-- function(p) return p:Nick() end\n-- )\n``` \nReason: This does filter and map in a single numeric for loop, so it avoids the extra function-call overhead, iterator chaining, and intermediate tables that fn combinators usually add.", "durationMs": 5600, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "ead0a98b3ff6a388", "startedAt": "2026-07-10T17:26:28.418Z", "completedAt": "2026-07-10T17:26:34.020Z", "answerBytes": 653, "answerChars": 653, "exitCode": 0, "usage": { "promptTokens": 76, "completionTokens": 164, "totalTokens": 240, "source": "estimated" } }, { "fixtureId": "gmod.perf.player-cache-hooks.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal cachedPlayers = {}\n\nlocal function refreshPlayers()\n cachedPlayers = player.GetAll()\nend\n\nhook.Add(\"PlayerInitialSpawn\", \"RefreshCachedPlayers\", refreshPlayers)\nhook.Add(\"PlayerDisconnected\", \"RefreshCachedPlayers\", refreshPlayers)\n\nhook.Add(\"Tick\", \"ProcessCachedPlayers\", function()\n for i = 1, #cachedPlayers do\n local ply = cachedPlayers[i]\n if IsValid(ply) then\n -- work with ply here\n end\n end\nend)\n\nrefreshPlayers()\n``` \nReason: `player.GetAll()` allocates and walks the player list; caching it and refreshing only on join/leave removes that repeated per-tick work.", "durationMs": 6809, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "a834817afe9e6c25", "startedAt": "2026-07-10T17:26:28.742Z", "completedAt": "2026-07-10T17:26:35.552Z", "answerBytes": 626, "answerChars": 626, "exitCode": 0, "usage": { "promptTokens": 88, "completionTokens": 157, "totalTokens": 245, "source": "estimated" } }, { "fixtureId": "gmod.perf.ply-table-index.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal tempData = setmetatable({}, { __mode = \"k\" }) -- weak keys so player entries can vanish\n\n-- write\nfunction SetTempValue(ply, value)\n tempData[ply] = value\nend\n\n-- read\nfunction GetTempValue(ply)\n return tempData[ply]\nend\n``` \nReason: a single shared lookup table avoids creating extra fields on each player object and gives you a tight `tempData[ply]` read path for hot per-player state.", "durationMs": 5033, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f18cc80df8999147", "startedAt": "2026-07-10T17:26:28.769Z", "completedAt": "2026-07-10T17:26:33.804Z", "answerBytes": 406, "answerChars": 406, "exitCode": 0, "usage": { "promptTokens": 69, "completionTokens": 102, "totalTokens": 171, "source": "estimated" } }, { "fixtureId": "gmod.perf.seq-tables.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain 0 through 1 candidate loops.", "finalResponse": "```lua\nlocal array = { \"a\", \"b\", \"c\" }\nlocal map = { first = \"a\", second = \"b\", third = \"c\" }\n\n-- Faster for fixed-order iteration:\nfor i = 1, #array do\n local value = array[i]\nend\n\n-- Slower for this use case:\nfor key, value in pairs(map) do\nend\n```\nReason: Sequential number-indexed arrays are faster to iterate than string-keyed maps because Lua can traverse the array part directly, while string-keyed tables use hash lookups.", "durationMs": 4386, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "4778b0220c3f91a3", "startedAt": "2026-07-10T17:26:28.994Z", "completedAt": "2026-07-10T17:26:33.383Z", "answerBytes": 433, "answerChars": 433, "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14549, "completionTokens": 172, "totalTokens": 14721, "reasoningTokens": 27, "cachedTokens": 13824 } }, { "fixtureId": "gmod.perf.setdrawcolor-split.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nsurface.SetDrawColor(255, 255, 255, 255)\n```\nReason: Passing raw r,g,b,a numbers avoids constructing or unpacking a `Color` object, which is the lighter choice for a hot draw path.", "durationMs": 4439, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "d110fde8a37bbe59", "startedAt": "2026-07-10T17:26:29.039Z", "completedAt": "2026-07-10T17:26:33.480Z", "answerBytes": 187, "answerChars": 187, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 47, "totalTokens": 118, "source": "estimated" } }, { "fixtureId": "gmod.perf.sqlite-vs-mysql.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nsql.Query(\"SELECT * FROM player_data WHERE steamid = 'STEAM_0:1:12345'\")\n```\nReason: Local SQLite is typically lower latency for a single Garry's Mod server because it avoids network round trips and uses an on-disk embedded database.", "durationMs": 3743, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "4a3f5d2a20d71ac9", "startedAt": "2026-07-10T17:26:29.235Z", "completedAt": "2026-07-10T17:26:32.979Z", "answerBytes": 240, "answerChars": 240, "exitCode": 0, "usage": { "promptTokens": 100, "completionTokens": 60, "totalTokens": 160, "source": "estimated" } }, { "fixtureId": "gmod.perf.steamid-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain exactly one fenced code block followed by a reason.", "finalResponse": "```lua\nlocal steamid_cache = setmetatable({}, { __mode = \"k\" })\n\nlocal function GetSteamIDCached(ply)\n local steamid = steamid_cache[ply]\n if steamid == nil then\n steamid = ply:SteamID()\n steamid_cache[ply] = steamid\n end\n return steamid\nend\n\nhook.Add(\"PlayerDisconnected\", \"ClearSteamIDCache\", function(ply)\n steamid_cache[ply] = nil\nend)\n\n-- hot path usage:\n-- local sid = GetSteamIDCached(ply)\n```lua\nReason: Caches each player's SteamID once and reuses it on hot paths, avoiding repeated method calls while the weak-key table lets entries fall out when the player entity is no longer referenced.", "durationMs": 6269, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "e8a94aa454154982", "startedAt": "2026-07-10T17:26:29.385Z", "completedAt": "2026-07-10T17:26:35.656Z", "answerBytes": 628, "answerChars": 628, "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14541, "completionTokens": 204, "totalTokens": 14745, "reasoningTokens": 12, "cachedTokens": 13824 } }, { "fixtureId": "gmod.perf.string-table-concat.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain 0 through 0 candidate loops.", "finalResponse": "```lua\n-- Slow: repeated .. creates many intermediate strings\nlocal out = \"\"\nfor i = 1, 10000 do\n out = out .. \"piece\" .. i .. \",\"\nend\n\n-- Fast: collect parts, then join once\nlocal parts = {}\nfor i = 1, 10000 do\n parts[#parts + 1] = \"piece\"\n parts[#parts + 1] = tostring(i)\n parts[#parts + 1] = \",\"\nend\nlocal out2 = table.concat(parts)\n```\nReason: `table.concat` builds the final string in one pass, avoiding the repeated intermediate allocations and copies caused by `..` in a loop.", "durationMs": 5511, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "b0cf9e04ae7c5dd9", "startedAt": "2026-07-10T17:26:29.753Z", "completedAt": "2026-07-10T17:26:35.265Z", "answerBytes": 495, "answerChars": 495, "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14546, "completionTokens": 181, "totalTokens": 14727, "reasoningTokens": 8, "cachedTokens": 13824 } }, { "fixtureId": "gmod.perf.surface-drawrect.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nsurface.SetDrawColor(r, g, b, a)\nsurface.DrawRect(x, y, w, h)\n```\nReason: `surface.DrawRect` skips the rounded-box path and is the lighter direct rectangle draw for non-rounded fills.", "durationMs": 3987, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "696e8de769d9ad0e", "startedAt": "2026-07-10T17:26:30.591Z", "completedAt": "2026-07-10T17:26:34.579Z", "answerBytes": 190, "answerChars": 190, "exitCode": 0, "usage": { "promptTokens": 69, "completionTokens": 48, "totalTokens": 117, "source": "estimated" } }, { "fixtureId": "gmod.perf.surface-text.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal fontName = \"HudHintTextLarge\"\nlocal text = \"Hello world\"\nlocal x, y = 100, 100\n\nsurface.SetFont(fontName)\nsurface.SetTextColor(255, 255, 255, 255)\nsurface.SetTextPos(x, y)\nsurface.DrawText(text)\n``` \nReason: `surface.SetFont` + `surface.SetTextPos` + `surface.SetTextColor` + `surface.DrawText` is the lightest option here; it avoids the extra layout/shadow/outline work and higher-level wrapper overhead in `draw.DrawText` and especially `SimpleTextOutlined`.", "durationMs": 6090, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "95d9fe4918ee4197", "startedAt": "2026-07-10T17:26:31.062Z", "completedAt": "2026-07-10T17:26:37.153Z", "answerBytes": 473, "answerChars": 473, "exitCode": 0, "usage": { "promptTokens": 76, "completionTokens": 119, "totalTokens": 195, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-count-hash.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal n = #t\n```\nReason: `#` is the preferred operator for the length of a sequential array-like table in Garry's Mod Lua.", "durationMs": 3183, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3e9bbfc8452e05e3", "startedAt": "2026-07-10T17:26:31.222Z", "completedAt": "2026-07-10T17:26:34.406Z", "answerBytes": 129, "answerChars": 129, "exitCode": 0, "usage": { "promptTokens": 62, "completionTokens": 33, "totalTokens": 95, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-empty.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\n-- Fastest when you only need to discard the old contents and do not need\n-- other references to see the cleared table.\nt = {}\n\n-- Use table.Empty(t) only when you must keep the same table object\n-- for existing references.\n``` \nReason: `t = {}` usually wins for repeatedly discarding a large table because it avoids walking and clearing every slot; `table.Empty(t)` only helps when you must preserve the original table reference.", "durationMs": 7181, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3a6344977f5bd475", "startedAt": "2026-07-10T17:26:31.642Z", "completedAt": "2026-07-10T17:26:38.824Z", "answerBytes": 437, "answerChars": 437, "exitCode": 0, "usage": { "promptTokens": 73, "completionTokens": 110, "totalTokens": 183, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-hasvalue.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal set = {\n apple = true,\n banana = true,\n cherry = true,\n}\n\nif set[item] then\n -- item exists in the set\nend\n```\nReason: A keyed lookup is O(1) on average, while `table.HasValue` scans a list linearly and gets slower as the list grows.", "durationMs": 5563, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "7d064428a873fdfc", "startedAt": "2026-07-10T17:26:31.798Z", "completedAt": "2026-07-10T17:26:37.362Z", "answerBytes": 258, "answerChars": 258, "exitCode": 0, "usage": { "promptTokens": 67, "completionTokens": 65, "totalTokens": 132, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-literal.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\n-- Faster / smaller: use the table constructor when fields are known up front.\nlocal t = { x = 1, y = 2, z = 3 }\n\n-- Slower: create empty table, then assign fields.\nlocal u = {}\nu.x = 1\nu.y = 2\nu.z = 3\n```\nReason: The constructor does the allocation and field initialization in one step, while separate assignments add extra writes and lookups; for a small known table, the constructor is the faster choice.", "durationMs": 4939, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f95a8fc610d720ad", "startedAt": "2026-07-10T17:26:32.342Z", "completedAt": "2026-07-10T17:26:37.283Z", "answerBytes": 414, "answerChars": 414, "exitCode": 0, "usage": { "promptTokens": 70, "completionTokens": 104, "totalTokens": 174, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-random.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal t = {\"a\", \"b\", \"c\", \"d\"}\nlocal value = t[math.random(#t)]\n``` \nReason: For a sequential table, indexing directly with `math.random(#t)` avoids `table.Random`'s extra table iteration and is faster.", "durationMs": 3451, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f3e090701953d4a9", "startedAt": "2026-07-10T17:26:32.799Z", "completedAt": "2026-07-10T17:26:36.252Z", "answerBytes": 209, "answerChars": 209, "exitCode": 0, "usage": { "promptTokens": 66, "completionTokens": 53, "totalTokens": 119, "source": "estimated" } }, { "fixtureId": "gmod.perf.x-times-x.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal y = x * x\n```\nReason: `x * x` is the direct square operation and is faster and clearer than calling `math.pow` or `x ^ 2` in a tight Lua loop.", "durationMs": 2983, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "79d48008f6b5174f", "startedAt": "2026-07-10T17:26:32.980Z", "completedAt": "2026-07-10T17:26:35.964Z", "answerBytes": 155, "answerChars": 155, "exitCode": 0, "usage": { "promptTokens": 58, "completionTokens": 39, "totalTokens": 97, "source": "estimated" } }, { "fixtureId": "gmod.player-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attemptIndex": 1, "status": "incorrect", "detail": "The recommendation changes the set of visited players.", "finalResponse": "```lua\nfor _, ply in ipairs(player.GetHumans()) do\n -- your code here\nend\n```\nReason: `player.GetHumans()` returns the currently connected human players, and `ipairs` is the idiomatic way to iterate that list in order.", "durationMs": 3564, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 2, "rubricVersion": "4", "promptHash": "fe3d7469535a0ca2", "startedAt": "2026-07-10T17:26:33.384Z", "completedAt": "2026-07-10T17:26:36.951Z", "answerBytes": 221, "answerChars": 221, "exitCode": 0, "usage": { "promptTokens": 80, "completionTokens": 56, "totalTokens": 136, "source": "estimated" } }, { "fixtureId": "gmod.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor ent in ents.Iterator() do\n if IsValid(ent) and ent:GetClass() == \"prop_physics\" then\n -- do something with ent\n end\nend\n```\nReason: `ents.Iterator()` yields entities lazily, so it scans the map without first materializing a full entity table.", "durationMs": 6304, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "3990c92253db85bb", "startedAt": "2026-07-10T17:26:33.481Z", "completedAt": "2026-07-10T17:26:39.786Z", "answerBytes": 266, "answerChars": 266, "exitCode": 0, "usage": { "promptTokens": 72, "completionTokens": 67, "totalTokens": 139, "source": "estimated" } }, { "fixtureId": "gmod.hook-add.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Registers a named hook with hook.Add.", "finalResponse": "```lua\nhook.Add(\"Think\", \"MyAddon.Think\", function()\n -- Think callback\nend)\n```\nReason: `hook.Add` registers the callback on Garry's Mod's `Think` hook under the identifier `MyAddon.Think`.", "durationMs": 4727, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "12d2222fdb7d6b90", "startedAt": "2026-07-10T17:26:33.634Z", "completedAt": "2026-07-10T17:26:38.363Z", "answerBytes": 193, "answerChars": 193, "exitCode": 0, "usage": { "promptTokens": 78, "completionTokens": 49, "totalTokens": 127, "source": "estimated" } }, { "fixtureId": "gmod.isvalid.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Uses IsValid for entity/object validity.", "finalResponse": "```lua\nif not IsValid(ent) then return end\n```\nReason: `IsValid(ent)` is the standard Garry's Mod check for whether an entity reference still points to a live, usable entity before calling methods on it.", "durationMs": 5499, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "35527312e59dfc13", "startedAt": "2026-07-10T17:26:33.804Z", "completedAt": "2026-07-10T17:26:39.305Z", "answerBytes": 203, "answerChars": 203, "exitCode": 0, "usage": { "promptTokens": 74, "completionTokens": 51, "totalTokens": 125, "source": "estimated" } }, { "fixtureId": "gmod.net-receivers.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nnet.Receive(\"MyAddon.Sync\", function(len, ply)\n -- handle the message here\nend)\n```\nReason: `net.Receive` is the Garry's Mod API for registering a handler for an incoming net message by name.", "durationMs": 5135, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "24d54ee8366753a6", "startedAt": "2026-07-10T17:26:34.020Z", "completedAt": "2026-07-10T17:26:39.157Z", "answerBytes": 201, "answerChars": 201, "exitCode": 0, "usage": { "promptTokens": 64, "completionTokens": 51, "totalTokens": 115, "source": "estimated" } }, { "fixtureId": "gmod.perf.angle-zero.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain exactly one fenced code block followed by a reason.", "finalResponse": "```lua\n-- Slow: allocates a new Angle every call.\nlocal function makeZeroAngle_slow()\n return Angle()\nend\n\n-- Faster in a hot path: reuse one scratch Angle and reset it.\n-- Only use this when the caller consumes it immediately and does not keep it.\nlocal ZERO_ANGLE = Angle()\n\nlocal function makeZeroAngle_fast()\n ZERO_ANGLE:Zero()\n return ZERO_ANGLE\nend\n```", "durationMs": 16956, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f6bc97bf98280d62", "startedAt": "2026-07-10T17:26:34.407Z", "completedAt": "2026-07-10T17:26:51.364Z", "answerBytes": 367, "answerChars": 367, "exitCode": 0, "usage": { "source": "provider", "promptTokens": 16630, "completionTokens": 877, "totalTokens": 17507, "reasoningTokens": 762, "cachedTokens": 1536 } }, { "fixtureId": "gmod.perf.config-var.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal myaddon_color = Color(255, 0, 0)\n\nlocal c = myaddon_color\n```\nReason: A flat local is faster to read than `myaddon.config.color` because it avoids repeated table lookups.", "durationMs": 6613, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "bcc4ff8eb162b699", "startedAt": "2026-07-10T17:26:34.488Z", "completedAt": "2026-07-10T17:26:41.103Z", "answerBytes": 183, "answerChars": 183, "exitCode": 0, "usage": { "promptTokens": 82, "completionTokens": 46, "totalTokens": 128, "source": "estimated" } }, { "fixtureId": "gmod.perf.darkrpvar.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal jobName = ply:getDarkRPVar(\"job\")\n```\nReason: DarkRP exposes the job label through `getDarkRPVar(\"job\")`; `NWString(\"job\")` is an internal networked variable and is not the preferred API.", "durationMs": 5505, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "02a37daa699cded3", "startedAt": "2026-07-10T17:26:34.581Z", "completedAt": "2026-07-10T17:26:40.090Z", "answerBytes": 200, "answerChars": 200, "exitCode": 0, "usage": { "promptTokens": 66, "completionTokens": 50, "totalTokens": 116, "source": "estimated" } }, { "fixtureId": "gmod.perf.disttosqr.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nif vec1:DistToSqr(vec2) <= threshold * threshold then\n -- within range\nend\n```\nReason: `DistToSqr` avoids the square root and is the right choice when you only need a distance threshold check.", "durationMs": 3880, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "a5772470c6859cdc", "startedAt": "2026-07-10T17:26:35.266Z", "completedAt": "2026-07-10T17:26:39.148Z", "answerBytes": 202, "answerChars": 202, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 51, "totalTokens": 122, "source": "estimated" } }, { "fixtureId": "gmod.perf.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor ent in ents.Iterator() do\n -- visit every entity here\nend\n```\nReason: `ents.Iterator()` is the wiki-recommended efficient iterator for walking all entities without building a full table first.", "durationMs": 8815, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "306d3f1d05f28246", "startedAt": "2026-07-10T17:26:35.553Z", "completedAt": "2026-07-10T17:26:44.369Z", "answerBytes": 206, "answerChars": 206, "exitCode": 0, "usage": { "promptTokens": 66, "completionTokens": 52, "totalTokens": 118, "source": "estimated" } }, { "fixtureId": "gmod.perf.find-ents-near.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain 0 through 0 candidate loops.", "finalResponse": "```lua\nlocal pos = ply:GetPos()\nlocal radius = 256\nlocal entsNear = ents.FindInSphere(pos, radius)\n\nfor _, ent in ipairs(entsNear) do\n -- process ent\nend\n``` \nReason: `ents.FindInSphere` is a fast built-in broadphase query for “near a player” checks and is usually the simplest performant choice when you do not need the extra cost of `FindInPVS`.", "durationMs": 7801, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "07c583fc3d9fe09f", "startedAt": "2026-07-10T17:26:35.656Z", "completedAt": "2026-07-10T17:26:43.458Z", "answerBytes": 354, "answerChars": 350, "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14560, "completionTokens": 322, "totalTokens": 14882, "reasoningTokens": 208, "cachedTokens": 13824 } }, { "fixtureId": "gmod.perf.find-players-near.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal function PlayersInSphere(pos, radius)\n local radiusSqr = radius * radius\n local out = {}\n\n for _, ply in ipairs(player.GetAll()) do\n if ply:Alive() and ply:GetPos():DistToSqr(pos) <= radiusSqr then\n out[#out + 1] = ply\n end\n end\n\n return out\nend\n``` \nReason: `player.GetAll()` skips scanning every entity, and `DistToSqr` avoids the square root cost of `Distance`, which is faster when you only care about players.", "durationMs": 14581, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "d139ba2e8ee38a2f", "startedAt": "2026-07-10T17:26:35.965Z", "completedAt": "2026-07-10T17:26:50.547Z", "answerBytes": 467, "answerChars": 467, "exitCode": 0, "usage": { "promptTokens": 79, "completionTokens": 117, "totalTokens": 196, "source": "estimated" } }, { "fixtureId": "gmod.perf.for-vs-while.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor i = 1, n do\n -- tight counting loop\nend\n```\nReason: A numeric `for` loop is generally faster than `while` in Garry's Mod Lua because it is a built-in counted loop with less per-iteration overhead.", "durationMs": 17370, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "b3401a190a8aecf0", "startedAt": "2026-07-10T17:26:36.253Z", "completedAt": "2026-07-10T17:26:53.624Z", "answerBytes": 210, "answerChars": 210, "exitCode": 0, "usage": { "promptTokens": 65, "completionTokens": 53, "totalTokens": 118, "source": "estimated" } }, { "fixtureId": "gmod.perf.hook-once.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal players = player.GetAll()\n\nhook.Run(\"MyCustomTick\", players)\n```\nReason: one hook dispatch per tick is cheaper than one hook dispatch per player, because it avoids repeated hook lookup, argument packing, and Lua/C boundary overhead.", "durationMs": 5565, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "ed1d3e09249fc8bb", "startedAt": "2026-07-10T17:26:36.358Z", "completedAt": "2026-07-10T17:26:41.924Z", "answerBytes": 245, "answerChars": 245, "exitCode": 0, "usage": { "promptTokens": 84, "completionTokens": 62, "totalTokens": 146, "source": "estimated" } }, { "fixtureId": "gmod.perf.hudpaint-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nhook.Add(\"HUDPaint\", \"MyHUD\", function()\n local lp = LocalPlayer()\n if not IsValid(lp) then return end\n\n local hp = lp:Health()\n\n draw.SimpleText(hp, \"DermaDefault\", 20, 20, color_white)\n draw.SimpleText(\"HP: \" .. hp, \"DermaDefault\", 20, 40, color_white)\n draw.SimpleText(math.Clamp(hp, 0, 100), \"DermaDefault\", 20, 60, color_white)\nend)\n```\nReason: `LocalPlayer():Health()` is computed once per frame and reused, so the hook avoids repeating the same lookup and method call three times.", "durationMs": 8267, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3def450f3158ceff", "startedAt": "2026-07-10T17:26:36.952Z", "completedAt": "2026-07-10T17:26:45.221Z", "answerBytes": 512, "answerChars": 512, "exitCode": 0, "usage": { "promptTokens": 76, "completionTokens": 128, "totalTokens": 204, "source": "estimated" } }, { "fixtureId": "gmod.perf.local-color.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal WHITE = Color(255, 255, 255)\n\nfor i = 1, n do\n surface.SetDrawColor(WHITE)\n -- paint work\nend\n```\nReason: `Color(255, 255, 255)` allocates or looks up a color object each pass; hoisting one shared white color avoids repeated work in the hot loop.", "durationMs": 4921, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "16a7bed1b931d467", "startedAt": "2026-07-10T17:26:37.153Z", "completedAt": "2026-07-10T17:26:42.076Z", "answerBytes": 265, "answerChars": 265, "exitCode": 0, "usage": { "promptTokens": 63, "completionTokens": 67, "totalTokens": 130, "source": "estimated" } }, { "fixtureId": "gmod.perf.local-vs-global.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal sin = math.sin\n\nfor i = 1, n do\n local v = sin(i)\n -- use v\nend\n```\nReason: Caching `math.sin` in a local avoids repeated table lookups in a hot loop.", "durationMs": 4847, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "c116d8c651190647", "startedAt": "2026-07-10T17:26:37.284Z", "completedAt": "2026-07-10T17:26:42.132Z", "answerBytes": 169, "answerChars": 169, "exitCode": 0, "usage": { "promptTokens": 62, "completionTokens": 43, "totalTokens": 105, "source": "estimated" } }, { "fixtureId": "gmod.perf.localplayer-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal LocalPlayer = LocalPlayer\n\nhook.Add(\"HUDPaint\", \"MyFastCheck\", function()\n local lp = LocalPlayer()\n if not IsValid(lp) then return end\n\n -- use lp here\nend)\n``` \nReason: `LocalPlayer` is resolved once into a local upvalue, so repeated calls skip the global table lookup.", "durationMs": 8949, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "7398105af47b079e", "startedAt": "2026-07-10T17:26:37.363Z", "completedAt": "2026-07-10T17:26:46.316Z", "answerBytes": 293, "answerChars": 293, "exitCode": 0, "usage": { "promptTokens": 70, "completionTokens": 74, "totalTokens": 144, "source": "estimated" } }, { "fixtureId": "gmod.perf.looking-at.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "partial", "detail": "Matched a partial-answer pattern.", "finalResponse": "```lua\nlocal LOOK_DOT = 0.94 -- about a 20 degree cone; tune as needed\n\nlocal function ArePlayersLookingAtEachOther(plyA, plyB)\n if not IsValid(plyA) or not IsValid(plyB) then return false end\n\n local aPos = plyA:EyePos()\n local bPos = plyB:EyePos()\n\n local aToB = bPos - aPos\n local bToA = aPos - bPos\n\n if aToB:LengthSqr() == 0 or bToA:LengthSqr() == 0 then return false end\n\n aToB:Normalize()\n bToA:Normalize()\n\n return plyA:GetAimVector():Dot(aToB) >= LOOK_DOT\n and plyB:GetAimVector():Dot(bToA) >= LOOK_DOT\nend\n``` \nReason: Dot products give the same facing test as an angle check without `math.acos`, so this is cheaper and fast for per-tick use.", "durationMs": 21420, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "20b22f678a4b08d7", "startedAt": "2026-07-10T17:26:38.363Z", "completedAt": "2026-07-10T17:26:59.785Z", "answerBytes": 685, "answerChars": 685, "exitCode": 0, "usage": { "promptTokens": 91, "completionTokens": 172, "totalTokens": 263, "source": "estimated" } }, { "fixtureId": "gmod.perf.meta-vs-arg.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nSomeHelper(ply)\n```\nReason: `SomeHelper(ply)` is generally faster for a helper you control because it avoids the method lookup and colon-call sugar overhead; the difference is usually small, but the direct function call is the cheaper path.", "durationMs": 9692, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "de332f7356830dfa", "startedAt": "2026-07-10T17:26:38.824Z", "completedAt": "2026-07-10T17:26:48.517Z", "answerBytes": 247, "answerChars": 247, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 62, "totalTokens": 133, "source": "estimated" } }, { "fixtureId": "gmod.perf.msg-vs-print.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nMsgN(\"debug: \", value)\n```\nReason: `MsgN` writes straight to the console with less formatting overhead than `print`.", "durationMs": 8732, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "6cbb4e4b2f19dd24", "startedAt": "2026-07-10T17:26:39.148Z", "completedAt": "2026-07-10T17:26:47.881Z", "answerBytes": 123, "answerChars": 123, "exitCode": 0, "usage": { "promptTokens": 63, "completionTokens": 31, "totalTokens": 94, "source": "estimated" } }, { "fixtureId": "gmod.perf.pairs-ipairs-for.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor i = 1, #t do\n local v = t[i]\n -- use v\nend\n```\nReason: `numeric for` is typically fastest for a large sequential array-like table because it avoids iterator overhead; `ipairs` is slower and `pairs` is the least suitable for dense arrays.", "durationMs": 4119, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "de5e4875fcf3e600", "startedAt": "2026-07-10T17:26:39.158Z", "completedAt": "2026-07-10T17:26:43.278Z", "answerBytes": 254, "answerChars": 254, "exitCode": 0, "usage": { "promptTokens": 73, "completionTokens": 64, "totalTokens": 137, "source": "estimated" } }, { "fixtureId": "gmod.perf.plain-lua-vs-fn.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal function filter_map(list, pred, map)\n local out = {}\n local n = 0\n\n for i = 1, #list do\n local v = list[i]\n if pred(v, i) then\n n = n + 1\n out[n] = map(v, i)\n end\n end\n\n return out\nend\n\n-- Example: DarkRP-style list work, but fused into one pass.\nlocal names = filter_map(players.GetAll(), function(p)\n return IsValid(p) and p:Alive() and not p:isArrested()\nend, function(p)\n return p:Nick()\nend)\n``` \nReason: One numeric for-loop, no intermediate tables, no combinator chaining, and no `table.insert` overhead; that is less allocation and fewer Lua calls than a small fn pipeline.", "durationMs": 31117, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "ead0a98b3ff6a388", "startedAt": "2026-07-10T17:26:39.306Z", "completedAt": "2026-07-10T17:27:10.424Z", "answerBytes": 620, "answerChars": 620, "exitCode": 0, "usage": { "promptTokens": 76, "completionTokens": 155, "totalTokens": 231, "source": "estimated" } }, { "fixtureId": "gmod.perf.player-cache-hooks.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal cachedPlayers = player.GetAll()\n\nhook.Add(\"PlayerInitialSpawn\", \"RefreshCachedPlayersJoin\", function(ply)\n cachedPlayers[#cachedPlayers + 1] = ply\nend)\n\nhook.Add(\"PlayerDisconnected\", \"RefreshCachedPlayersLeave\", function(ply)\n for i = #cachedPlayers, 1, -1 do\n if cachedPlayers[i] == ply then\n table.remove(cachedPlayers, i)\n break\n end\n end\nend)\n\nhook.Add(\"Tick\", \"UseCachedPlayers\", function()\n for i = 1, #cachedPlayers do\n local ply = cachedPlayers[i]\n -- work with ply\n end\nend)\n``` \nReason: `player.GetAll()` allocates and rebuilds the full player list every tick; caching it and updating only on join/leave removes that repeated work.", "durationMs": 6514, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "a834817afe9e6c25", "startedAt": "2026-07-10T17:26:39.787Z", "completedAt": "2026-07-10T17:26:46.302Z", "answerBytes": 720, "answerChars": 720, "exitCode": 0, "usage": { "promptTokens": 88, "completionTokens": 180, "totalTokens": 268, "source": "estimated" } }, { "fixtureId": "gmod.perf.ply-table-index.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal PlayerData = setmetatable({}, { __mode = \"k\" }) -- weak keys; entries vanish when player entity is gone\n\nlocal function GetPlayerData(ply)\n local data = PlayerData[ply]\n if data == nil then\n data = {}\n PlayerData[ply] = data\n end\n return data\nend\n\n-- write\nlocal data = GetPlayerData(ply)\ndata.tempScore = 123\n\n-- hot reads\nlocal score = GetPlayerData(ply).tempScore\n``` \nReason: a plain Lua table keyed by the player entity avoids repeated entity-field lookup, and hot code can keep the returned data table in a local for faster reads.", "durationMs": 13329, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f18cc80df8999147", "startedAt": "2026-07-10T17:26:40.090Z", "completedAt": "2026-07-10T17:26:53.421Z", "answerBytes": 575, "answerChars": 575, "exitCode": 0, "usage": { "promptTokens": 69, "completionTokens": 144, "totalTokens": 213, "source": "estimated" } }, { "fixtureId": "gmod.perf.seq-tables.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal values = {\n \"alpha\",\n \"beta\",\n \"gamma\",\n \"delta\",\n}\n\nfor i = 1, #values do\n local v = values[i]\n -- use v\nend\n``` \nReason: Sequential number-indexed arrays are faster to iterate than string-keyed maps because Lua can walk them with dense numeric access instead of hash lookups.", "durationMs": 6770, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "4778b0220c3f91a3", "startedAt": "2026-07-10T17:26:41.104Z", "completedAt": "2026-07-10T17:26:47.876Z", "answerBytes": 296, "answerChars": 296, "exitCode": 0, "usage": { "promptTokens": 79, "completionTokens": 74, "totalTokens": 153, "source": "estimated" } }, { "fixtureId": "gmod.perf.setdrawcolor-split.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nsurface.SetDrawColor(255, 128, 64, 255)\n```\nReason: Passing raw numbers avoids creating or unpacking a `Color` object on a hot draw path, so it is the faster call style.", "durationMs": 5005, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "d110fde8a37bbe59", "startedAt": "2026-07-10T17:26:41.195Z", "completedAt": "2026-07-10T17:26:46.201Z", "answerBytes": 176, "answerChars": 176, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 44, "totalTokens": 115, "source": "estimated" } }, { "fixtureId": "gmod.perf.sqlite-vs-mysql.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal result = sql.QueryValue(\"SELECT money FROM player_data WHERE steamid = \" .. sql.SQLStr(steamID64))\n```\nReason: Local SQLite is typically lower latency for a single GMod server because it avoids network round-trips and remote MySQL connection overhead.", "durationMs": 4530, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "4a3f5d2a20d71ac9", "startedAt": "2026-07-10T17:26:41.926Z", "completedAt": "2026-07-10T17:26:46.457Z", "answerBytes": 264, "answerChars": 264, "exitCode": 0, "usage": { "promptTokens": 100, "completionTokens": 66, "totalTokens": 166, "source": "estimated" } }, { "fixtureId": "gmod.perf.steamid-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal steamid_cache = setmetatable({}, { __mode = \"k\" }) -- weak keys: entries drop when the player entity disappears\n\nlocal function SteamIDCached(ply)\n local sid = steamid_cache[ply]\n if sid ~= nil then\n return sid\n end\n\n sid = ply:SteamID()\n steamid_cache[ply] = sid\n return sid\nend\n\n-- hot path usage\n-- local sid = SteamIDCached(ply)\n``` \nReason: `ply:SteamID()` is an expensive repeated lookup here, so a weak-key memoization table returns the same string after the first call per player and lets stale entries fall away automatically.", "durationMs": 31017, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "e8a94aa454154982", "startedAt": "2026-07-10T17:26:42.077Z", "completedAt": "2026-07-10T17:27:13.095Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14543, "completionTokens": 522, "totalTokens": 15065, "reasoningTokens": 353, "cachedTokens": 13824 }, "answerBytes": 573, "answerChars": 573 }, { "fixtureId": "gmod.perf.string-table-concat.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain 0 through 0 candidate loops.", "finalResponse": "```lua\n-- Slow: repeated .. rebuilds the string over and over.\nlocal out = \"\"\nfor i = 1, 10000 do\n out = out .. \"piece\" .. i .. \",\"\nend\n\n-- Fast: collect parts, then join once.\nlocal parts = {}\nfor i = 1, 10000 do\n parts[#parts + 1] = \"piece\"\n parts[#parts + 1] = i\n parts[#parts + 1] = \",\"\nend\n\nlocal out_fast = table.concat(parts)\n``` \nReason: `..` copies the full string each time, so it gets slower as the string grows; `table.concat` builds once from parts and is much faster for thousands of pieces.", "durationMs": 13544, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "b0cf9e04ae7c5dd9", "startedAt": "2026-07-10T17:26:42.133Z", "completedAt": "2026-07-10T17:26:55.680Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14544, "completionTokens": 306, "totalTokens": 14850, "reasoningTokens": 121, "cachedTokens": 0 }, "answerBytes": 517, "answerChars": 517 }, { "fixtureId": "gmod.perf.surface-drawrect.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nsurface.SetDrawColor(color)\nsurface.DrawRect(x, y, w, h)\n```\nReason: `draw.RoundedBox(0, ...)` does extra rounded-box plumbing; `surface.DrawRect` is the direct filled-rectangle draw call and is a bit cheaper for plain rectangles.", "durationMs": 13269, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "696e8de769d9ad0e", "startedAt": "2026-07-10T17:26:43.279Z", "completedAt": "2026-07-10T17:26:56.549Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14546, "completionTokens": 250, "totalTokens": 14796, "reasoningTokens": 169, "cachedTokens": 13824 }, "answerBytes": 237, "answerChars": 237 }, { "fixtureId": "gmod.perf.surface-text.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal font = \"MyFont\"\nlocal text = \"Hello, world\"\nlocal x, y = 100, 100\nlocal color = color_white\n\nsurface.SetFont(font)\nsurface.SetTextColor(color)\nsurface.SetTextPos(x, y)\nsurface.DrawText(text)\n``` \nReason: `surface.SetFont` + `surface.SetTextPos` + `surface.SetTextColor` + `surface.DrawText` is the leanest path for a single line; `draw.DrawText` and especially `SimpleTextOutlined` do extra layout/alignment/outline work every frame.", "durationMs": 21969, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "95d9fe4918ee4197", "startedAt": "2026-07-10T17:26:43.459Z", "completedAt": "2026-07-10T17:27:05.430Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14547, "completionTokens": 427, "totalTokens": 14974, "reasoningTokens": 282, "cachedTokens": 0 }, "answerBytes": 446, "answerChars": 446 }, { "fixtureId": "gmod.perf.table-count-hash.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal len = #t\n```\nReason: `#` is the standard Lua length operator for sequential array-like tables.", "durationMs": 45198, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3e9bbfc8452e05e3", "startedAt": "2026-07-10T17:26:44.370Z", "completedAt": "2026-07-10T17:27:29.569Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14537, "completionTokens": 152, "totalTokens": 14689, "reasoningTokens": 107, "cachedTokens": 13824 }, "answerBytes": 107, "answerChars": 107 }, { "fixtureId": "gmod.perf.table-empty.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\n-- Faster when you can replace the reference:\nt = {}\n\n-- Use table.Empty(t) only when you must keep the same table object.\n``` \nReason: `t = {}` is generally faster because it allocates a new empty table instead of clearing every key in the old one; `table.Empty(t)` mutates the existing table and is typically slower for large tables.", "durationMs": 15481, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3a6344977f5bd475", "startedAt": "2026-07-10T17:26:45.222Z", "completedAt": "2026-07-10T17:27:00.705Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14548, "completionTokens": 242, "totalTokens": 14790, "reasoningTokens": 143, "cachedTokens": 13824 }, "answerBytes": 342, "answerChars": 342 }, { "fixtureId": "gmod.perf.table-hasvalue.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal allowed = {\n [\"weapon_pistol\"] = true,\n [\"weapon_smg1\"] = true,\n [\"weapon_ar2\"] = true,\n}\n\nif allowed[weaponName] then\n -- found\nend\n```\nReason: A keyed lookup is O(1) average, while `table.HasValue` scans the list linearly and gets slower as the list grows.", "durationMs": 19098, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "7d064428a873fdfc", "startedAt": "2026-07-10T17:26:46.202Z", "completedAt": "2026-07-10T17:27:05.301Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14544, "completionTokens": 181, "totalTokens": 14725, "reasoningTokens": 72, "cachedTokens": 13824 }, "answerBytes": 283, "answerChars": 283 }, { "fixtureId": "gmod.perf.table-literal.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal t = {\n x = 1,\n y = 2,\n z = 3,\n}\n``` \nReason: The constructor is faster because it builds the table with its fields in one step, instead of creating an empty table and doing extra assignments afterward.", "durationMs": 12602, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f95a8fc610d720ad", "startedAt": "2026-07-10T17:26:46.303Z", "completedAt": "2026-07-10T17:26:58.906Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14545, "completionTokens": 204, "totalTokens": 14749, "reasoningTokens": 124, "cachedTokens": 13824 }, "answerBytes": 223, "answerChars": 223 }, { "fixtureId": "gmod.perf.table-random.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal idx = math.random(#seq)\nlocal value = seq[idx]\n``` \nReason: For a sequential array, direct indexing with `math.random(#seq)` avoids the extra work `table.Random` does to scan/handle generic table keys.", "durationMs": 14617, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f3e090701953d4a9", "startedAt": "2026-07-10T17:26:46.316Z", "completedAt": "2026-07-10T17:27:00.936Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14541, "completionTokens": 188, "totalTokens": 14729, "reasoningTokens": 117, "cachedTokens": 0 }, "answerBytes": 214, "answerChars": 214 }, { "fixtureId": "gmod.perf.x-times-x.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal y = x * x\n```\nReason: `x * x` avoids the function call and overhead of `math.pow` in a tight loop.", "durationMs": 13175, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "79d48008f6b5174f", "startedAt": "2026-07-10T17:26:46.458Z", "completedAt": "2026-07-10T17:26:59.635Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14538, "completionTokens": 156, "totalTokens": 14694, "reasoningTokens": 104, "cachedTokens": 13824 }, "answerBytes": 111, "answerChars": 111 }, { "fixtureId": "gmod.player-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attemptIndex": 1, "status": "partial", "detail": "Uses ipairs but still allocates via player.GetAll() each time.", "finalResponse": "```lua\nfor _, ply in ipairs(player.GetAll()) do\n -- visit each connected player here\nend\n```\nReason: `player.GetAll()` returns the current players as an array-like table, so `ipairs` is the direct loop for visiting each connected player.", "durationMs": 11027, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 2, "rubricVersion": "4", "promptHash": "fe3d7469535a0ca2", "startedAt": "2026-07-10T17:26:47.877Z", "completedAt": "2026-07-10T17:26:58.905Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14550, "completionTokens": 119, "totalTokens": 14669, "reasoningTokens": 44, "cachedTokens": 13824 }, "answerBytes": 240, "answerChars": 240 }, { "fixtureId": "gmod.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor _, ent in ipairs(ents.FindByClass(\"prop_physics\")) do\n -- visit each prop_physics here\nend\n```\nReason: `ents.FindByClass` returns only entities of the requested class, so it avoids building a table of every entity in the map first.", "durationMs": 16615, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "3990c92253db85bb", "startedAt": "2026-07-10T17:26:47.882Z", "completedAt": "2026-07-10T17:27:04.498Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14545, "completionTokens": 445, "totalTokens": 14990, "reasoningTokens": 364, "cachedTokens": 0 }, "answerBytes": 245, "answerChars": 245 }, { "fixtureId": "gmod.hook-add.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Registers a named hook with hook.Add.", "finalResponse": "```lua\nhook.Add(\"Think\", \"MyAddon.Think\", function()\nend)\n```\nReason: `hook.Add` registers a function on Garry's Mod's `Think` hook, and `MyAddon.Think` is the hook identifier.", "durationMs": 12641, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "12d2222fdb7d6b90", "startedAt": "2026-07-10T17:26:48.518Z", "completedAt": "2026-07-10T17:27:01.160Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14549, "completionTokens": 167, "totalTokens": 14716, "reasoningTokens": 96, "cachedTokens": 0 }, "answerBytes": 176, "answerChars": 176 }, { "fixtureId": "gmod.isvalid.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Uses IsValid for entity/object validity.", "finalResponse": "```lua\nif IsValid(ent) then\n -- safe to call methods on ent here\nend\n```\nReason: `IsValid(ent)` is the standard Garry's Mod check for a live entity reference before using it.", "durationMs": 11576, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "35527312e59dfc13", "startedAt": "2026-07-10T17:26:50.548Z", "completedAt": "2026-07-10T17:27:02.125Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14542, "completionTokens": 174, "totalTokens": 14716, "reasoningTokens": 109, "cachedTokens": 0 }, "answerBytes": 177, "answerChars": 177 }, { "fixtureId": "gmod.net-receivers.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nnet.Receive(\"MyAddon.Sync\", function(len, ply)\n -- handle the message here\nend)\n```\nReason: `net.Receive` registers the callback that runs when the `MyAddon.Sync` net message arrives.", "durationMs": 12885, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "24d54ee8366753a6", "startedAt": "2026-07-10T17:26:51.365Z", "completedAt": "2026-07-10T17:27:04.252Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14535, "completionTokens": 220, "totalTokens": 14755, "reasoningTokens": 153, "cachedTokens": 0 }, "answerBytes": 193, "answerChars": 193 }, { "fixtureId": "gmod.perf.angle-zero.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "incorrect", "detail": "Did not match a fixture answer pattern.", "finalResponse": "```lua\n-- Fast: reuse the built-in zero angle instead of allocating a new one each time.\nlocal ZERO_ANGLE = angle_zero\n\n-- Example hot-path use:\n-- someFunc(ZERO_ANGLE)\n``` \nReason: `angle_zero` is a shared zeroed `Angle` value, so reusing it avoids repeated `Angle()` construction and allocation in hot paths.", "durationMs": 18039, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f6bc97bf98280d62", "startedAt": "2026-07-10T17:26:53.422Z", "completedAt": "2026-07-10T17:27:11.462Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14545, "completionTokens": 614, "totalTokens": 15159, "reasoningTokens": 516, "cachedTokens": 0 }, "answerBytes": 310, "answerChars": 310 }, { "fixtureId": "gmod.perf.config-var.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal addon_color = Color(255, 128, 0)\nlocal c = addon_color\n```\nReason: A local is faster to read than a nested table lookup or global access in Lua, so it is the best choice for a frequently read setting.", "durationMs": 15036, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "bcc4ff8eb162b699", "startedAt": "2026-07-10T17:26:53.625Z", "completedAt": "2026-07-10T17:27:08.663Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14550, "completionTokens": 415, "totalTokens": 14965, "reasoningTokens": 341, "cachedTokens": 13824 }, "answerBytes": 213, "answerChars": 213 }, { "fixtureId": "gmod.perf.darkrpvar.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal jobName = ply:getDarkRPVar(\"job\")\n```\nReason: DarkRP exposes the job through its own `DarkRPVar` API, so this is the preferred, supported way instead of reading the raw networked string.", "durationMs": 17008, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "02a37daa699cded3", "startedAt": "2026-07-10T17:26:55.681Z", "completedAt": "2026-07-10T17:27:12.691Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14546, "completionTokens": 538, "totalTokens": 15084, "reasoningTokens": 468, "cachedTokens": 13824 }, "answerBytes": 199, "answerChars": 199 }, { "fixtureId": "gmod.perf.disttosqr.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nif pos1:DistToSqr(pos2) <= threshold * threshold then\n -- within range\nend\n```\nReason: `DistToSqr` avoids the square root cost of `DistTo`, which is the right check when you only need to know whether two vectors are within a threshold.", "durationMs": 13234, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "a5772470c6859cdc", "startedAt": "2026-07-10T17:26:56.549Z", "completedAt": "2026-07-10T17:27:09.785Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14543, "completionTokens": 222, "totalTokens": 14765, "reasoningTokens": 139, "cachedTokens": 0 }, "answerBytes": 245, "answerChars": 245 }, { "fixtureId": "gmod.perf.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor ent in ents.Iterator() do\n -- visit each entity here\nend\n```\nReason: `ents.Iterator()` walks the entity list without allocating a full table, so it is the preferred efficient loop for visiting every entity.", "durationMs": 19485, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "306d3f1d05f28246", "startedAt": "2026-07-10T17:26:58.906Z", "completedAt": "2026-07-10T17:27:18.393Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14541, "completionTokens": 275, "totalTokens": 14816, "reasoningTokens": 208, "cachedTokens": 13824 }, "answerBytes": 220, "answerChars": 220 }, { "fixtureId": "gmod.perf.find-ents-near.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain 0 through 0 candidate loops.", "finalResponse": "```lua\nlocal function GetNearbyEntities(ply, radius)\n local pos = ply:GetPos()\n local out = ents.FindInSphere(pos, radius)\n\n local filtered = {}\n for i = 1, #out do\n local ent = out[i]\n if ent ~= ply and IsValid(ent) then\n filtered[#filtered + 1] = ent\n end\n end\n\n return filtered\nend\n```\nReason: `ents.FindInSphere` is a fast built-in broad-phase query for “nearby entities” and is commonly benchmarked as one of the better choices for radius checks, without using `FindInPVS` as the default.", "durationMs": 16975, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "07c583fc3d9fe09f", "startedAt": "2026-07-10T17:26:58.907Z", "completedAt": "2026-07-10T17:27:15.888Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14560, "completionTokens": 519, "totalTokens": 15079, "reasoningTokens": 353, "cachedTokens": 0 }, "answerBytes": 547, "answerChars": 543 }, { "fixtureId": "gmod.perf.find-players-near.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal function PlayersInSphere(pos, radius)\n local radiusSqr = radius * radius\n local found = {}\n\n for _, ply in ipairs(player.GetAll()) do\n if ply:GetPos():DistToSqr(pos) <= radiusSqr then\n found[#found + 1] = ply\n end\n end\n\n return found\nend\n``` \nReason: `player.GetAll()` only walks players, not every entity, and `DistToSqr` avoids a square root per check.", "durationMs": 15537, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "d139ba2e8ee38a2f", "startedAt": "2026-07-10T17:26:59.636Z", "completedAt": "2026-07-10T17:27:15.174Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14556, "completionTokens": 430, "totalTokens": 14986, "reasoningTokens": 298, "cachedTokens": 13824 }, "answerBytes": 407, "answerChars": 407 }, { "fixtureId": "gmod.perf.for-vs-while.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal sum = 0\nfor i = 1, n do\n sum = sum + i\nend\n```\nReason: Numeric `for` is generally faster in Lua for tight counted loops because the VM handles the counter and limit more efficiently than a manual `while` increment and compare.", "durationMs": 13576, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "b3401a190a8aecf0", "startedAt": "2026-07-10T17:26:59.786Z", "completedAt": "2026-07-10T17:27:13.363Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14547, "completionTokens": 296, "totalTokens": 14843, "reasoningTokens": 215, "cachedTokens": 13824 }, "answerBytes": 242, "answerChars": 242 }, { "fixtureId": "gmod.perf.hook-once.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nhook.Add(\"Think\", \"NotifyPlayersEachTick\", function()\n local players = player.GetAll()\n hook.Run(\"MyCustomPlayerTick\", players)\nend)\n``` \nReason: one `hook.Run` per tick is faster than one `hook.Run` per player because it avoids repeated hook dispatch and Lua/C boundary overhead.", "durationMs": 16582, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "ed1d3e09249fc8bb", "startedAt": "2026-07-10T17:27:00.706Z", "completedAt": "2026-07-10T17:27:17.290Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14555, "completionTokens": 490, "totalTokens": 15045, "reasoningTokens": 399, "cachedTokens": 0 }, "answerBytes": 293, "answerChars": 293 }, { "fixtureId": "gmod.perf.hudpaint-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nhook.Add(\"HUDPaint\", \"MyHealthHUD\", function()\n local lp = LocalPlayer()\n if not IsValid(lp) then return end\n\n local health = lp:Health()\n\n draw.SimpleText(health, \"DermaDefault\", 100, 100, color_white)\n draw.SimpleText(health, \"DermaDefault\", 100, 120, color_white)\n draw.SimpleText(health, \"DermaDefault\", 100, 140, color_white)\nend)\n```\nReason: Cache `LocalPlayer()` and `:Health()` once per frame so the three draw calls reuse the same value instead of repeating the lookup.", "durationMs": 14580, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3def450f3158ceff", "startedAt": "2026-07-10T17:27:00.936Z", "completedAt": "2026-07-10T17:27:15.518Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14552, "completionTokens": 285, "totalTokens": 14837, "reasoningTokens": 120, "cachedTokens": 13824 }, "answerBytes": 503, "answerChars": 503 }, { "fixtureId": "gmod.perf.local-color.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal WHITE = Color(255, 255, 255)\n\nhook.Add(\"HUDPaint\", \"MyPaintLoop\", function()\n for i = 1, 100 do\n surface.SetDrawColor(WHITE)\n surface.DrawRect(i, 10, 1, 1)\n end\nend)\n```\nReason: Reusing one prebuilt `Color` avoids allocating a new color object on every iteration.", "durationMs": 22799, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "16a7bed1b931d467", "startedAt": "2026-07-10T17:27:01.161Z", "completedAt": "2026-07-10T17:27:23.962Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14540, "completionTokens": 217, "totalTokens": 14757, "reasoningTokens": 104, "cachedTokens": 13824 }, "answerBytes": 296, "answerChars": 296 }, { "fixtureId": "gmod.perf.local-vs-global.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal sin = math.sin\n\nfor i = 1, 100000 do\n local y = sin(i)\n -- use y\nend\n``` \nReason: Caching `math.sin` in a local avoids repeated global/table lookups inside the hot loop.", "durationMs": 12297, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "c116d8c651190647", "startedAt": "2026-07-10T17:27:02.126Z", "completedAt": "2026-07-10T17:27:14.425Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14540, "completionTokens": 153, "totalTokens": 14693, "reasoningTokens": 76, "cachedTokens": 13824 }, "answerBytes": 188, "answerChars": 188 }, { "fixtureId": "gmod.perf.localplayer-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "partial", "detail": "Matched a partial-answer pattern.", "finalResponse": "```lua\nlocal LocalPlayer = LocalPlayer\nlocal IsValid = IsValid\n\nlocal cachedLP\n\nlocal function GetLocalPlayerCached()\n if not IsValid(cachedLP) then\n cachedLP = LocalPlayer()\n end\n\n return cachedLP\nend\n\n-- hot path\nlocal lp = GetLocalPlayerCached()\nif IsValid(lp) then\n -- reuse `lp` instead of calling LocalPlayer() repeatedly\nend\n``` \nReason: it turns repeated global/function lookups into one cached local reference that you reuse in hot code.", "durationMs": 28758, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "7398105af47b079e", "startedAt": "2026-07-10T17:27:02.184Z", "completedAt": "2026-07-10T17:27:30.945Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14547, "completionTokens": 1391, "totalTokens": 15938, "reasoningTokens": 1253, "cachedTokens": 13824 }, "answerBytes": 465, "answerChars": 465 }, { "fixtureId": "gmod.perf.looking-at.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "partial", "detail": "Matched a partial-answer pattern.", "finalResponse": "```lua\nlocal function RoughlyLookingAtEachOther(plyA, plyB, cosLimit)\n if not (IsValid(plyA) and IsValid(plyB)) then return false end\n\n cosLimit = cosLimit or 0.7 -- about 45 degrees; raise for a tighter check\n\n local aAim = plyA:GetAimVector() -- normalized\n local bAim = plyB:GetAimVector() -- normalized\n\n local aToB = (plyB:EyePos() - plyA:EyePos()):GetNormalized()\n local bToA = -aToB\n\n return aAim:Dot(aToB) >= cosLimit\n and bAim:Dot(bToA) >= cosLimit\nend\n``` \nReason: `Dot` plus a cosine threshold is fast and avoids `math.acos` while still testing whether each player is facing the other.", "durationMs": 21003, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "20b22f678a4b08d7", "startedAt": "2026-07-10T17:27:04.252Z", "completedAt": "2026-07-10T17:27:25.258Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14557, "completionTokens": 671, "totalTokens": 15228, "reasoningTokens": 458, "cachedTokens": 13824 }, "answerBytes": 623, "answerChars": 623 }, { "fixtureId": "gmod.perf.meta-vs-arg.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nSomeHelper(ply)\n```\nReason: for a helper you control, a plain function call avoids method lookup and the implicit `self` passing, so it is generally a bit faster than `ply:SomeHelper()`.", "durationMs": 16614, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "de332f7356830dfa", "startedAt": "2026-07-10T17:27:04.499Z", "completedAt": "2026-07-10T17:27:21.115Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14551, "completionTokens": 527, "totalTokens": 15078, "reasoningTokens": 461, "cachedTokens": 13824 }, "answerBytes": 193, "answerChars": 193 }, { "fixtureId": "gmod.perf.msg-vs-print.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nMsgN(\"hot path debug\")\n```\nReason: `MsgN` writes directly to the console with less overhead than `print` when you do not need Lua-style formatting or multiple-argument joining.", "durationMs": 18045, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "6cbb4e4b2f19dd24", "startedAt": "2026-07-10T17:27:05.301Z", "completedAt": "2026-07-10T17:27:23.347Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14544, "completionTokens": 387, "totalTokens": 14931, "reasoningTokens": 324, "cachedTokens": 0 }, "answerBytes": 183, "answerChars": 183 }, { "fixtureId": "gmod.perf.pairs-ipairs-for.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor i = 1, #t do\n local v = t[i]\n -- use v\nend\n```\nReason: `for i = 1, #t do` is typically fastest for a dense sequential array-like table because it avoids the iterator overhead of `pairs` and `ipairs`.", "durationMs": 11458, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "de5e4875fcf3e600", "startedAt": "2026-07-10T17:27:05.433Z", "completedAt": "2026-07-10T17:27:16.892Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14550, "completionTokens": 155, "totalTokens": 14705, "reasoningTokens": 69, "cachedTokens": 13824 }, "answerBytes": 216, "answerChars": 216 }, { "fixtureId": "gmod.perf.plain-lua-vs-fn.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\n-- Faster than chaining DarkRP fn filter/map combinators: one pass, no intermediates.\nlocal function filterMap(list, keep, map)\n local out = {}\n local n = 0\n\n for i = 1, #list do\n local v = list[i]\n if keep(v) then\n n = n + 1\n out[n] = map(v)\n end\n end\n\n return out\nend\n\n-- Example: keep active players and map to nicknames.\nlocal names = filterMap(player.GetAll(),\n function(p) return IsValid(p) and p:Alive() end,\n function(p) return p:Nick() end\n)\n``` \nReason: One tight loop with no combinator chaining, no extra closures per stage, and no intermediate lists, so it does less allocation and dispatch work.", "durationMs": 17296, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "ead0a98b3ff6a388", "startedAt": "2026-07-10T17:27:08.663Z", "completedAt": "2026-07-10T17:27:25.961Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14562, "completionTokens": 519, "totalTokens": 15081, "reasoningTokens": 310, "cachedTokens": 0 }, "answerBytes": 680, "answerChars": 680 }, { "fixtureId": "gmod.perf.player-cache-hooks.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal cachedPlayers = player.GetAll()\n\nlocal function RefreshPlayers()\n cachedPlayers = player.GetAll()\nend\n\nhook.Add(\"PlayerInitialSpawn\", \"RefreshCachedPlayersOnJoin\", RefreshPlayers)\nhook.Add(\"PlayerDisconnected\", \"RefreshCachedPlayersOnLeave\", RefreshPlayers)\n\nhook.Add(\"Think\", \"UseCachedPlayersInsteadOfPlayerGetAll\", function()\n for _, ply in ipairs(cachedPlayers) do\n if IsValid(ply) then\n -- work with ply\n end\n end\nend)\n``` \nReason: `player.GetAll()` allocates and repopulates a table every tick; caching once and refreshing only on join/leave removes that repeated per-frame work.", "durationMs": 17463, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "a834817afe9e6c25", "startedAt": "2026-07-10T17:27:09.786Z", "completedAt": "2026-07-10T17:27:27.250Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14559, "completionTokens": 599, "totalTokens": 15158, "reasoningTokens": 425, "cachedTokens": 13824 }, "answerBytes": 632, "answerChars": 632 }, { "fixtureId": "gmod.perf.ply-table-index.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal tempByPly = setmetatable({}, { __mode = \"k\" }) -- weak keys, auto-clears when player entity is gone\n\nlocal function SetTemp(ply, value)\n tempByPly[ply] = value\nend\n\nlocal function GetTemp(ply)\n return tempByPly[ply]\nend\n\n-- hot path\nlocal v = tempByPly[ply]\n``` \nReason: a weak-key side table avoids per-entity field lookup on `ply` and keeps reads as a direct table index by player reference.", "durationMs": 17249, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f18cc80df8999147", "startedAt": "2026-07-10T17:27:10.425Z", "completedAt": "2026-07-10T17:27:27.675Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14544, "completionTokens": 549, "totalTokens": 15093, "reasoningTokens": 406, "cachedTokens": 13824 }, "answerBytes": 412, "answerChars": 412 }, { "fixtureId": "gmod.perf.seq-tables.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\n-- Faster for iteration: sequential numeric indices\nlocal values = { \"alpha\", \"beta\", \"gamma\", \"delta\" }\n\nfor i = 1, #values do\n local v = values[i]\n -- use v\nend\n``` \nReason: Sequential number-indexed arrays are faster to iterate in Lua because they use direct array-style access, while string-keyed maps require hash lookup for each element.", "durationMs": 14693, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "4778b0220c3f91a3", "startedAt": "2026-07-10T17:27:11.463Z", "completedAt": "2026-07-10T17:27:26.157Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14547, "completionTokens": 342, "totalTokens": 14889, "reasoningTokens": 234, "cachedTokens": 13824 }, "answerBytes": 352, "answerChars": 352 }, { "fixtureId": "gmod.perf.setdrawcolor-split.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nsurface.SetDrawColor(255, 255, 255, 255)\n```\nReason: Separate numeric RGBA arguments avoid the extra Color-object handling and are the faster call style in a hot draw path.", "durationMs": 12966, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "d110fde8a37bbe59", "startedAt": "2026-07-10T17:27:12.692Z", "completedAt": "2026-07-10T17:27:25.660Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14552, "completionTokens": 235, "totalTokens": 14787, "reasoningTokens": 171, "cachedTokens": 13824 }, "answerBytes": 179, "answerChars": 179 }, { "fixtureId": "gmod.perf.sqlite-vs-mysql.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal row = sql.QueryRow(\"SELECT value FROM playerdata WHERE steamid = \" .. sql.SQLStr(steamid))\n```\nReason: For a single Garry's Mod server with no cross-server sync, local SQLite is typically lower latency because it stays in-process/on-disk, while MySQL adds network round trips and remote server overhead.", "durationMs": 16317, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "4a3f5d2a20d71ac9", "startedAt": "2026-07-10T17:27:13.096Z", "completedAt": "2026-07-10T17:27:29.414Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14568, "completionTokens": 458, "totalTokens": 15026, "reasoningTokens": 366, "cachedTokens": 13824 }, "answerBytes": 316, "answerChars": 316 }, { "fixtureId": "gmod.perf.steamid-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal steamid_cache = setmetatable({}, { __mode = \"k\" }) -- weak keys, no leak after players go away\n\nlocal function GetSteamIDFast(ply)\n local sid = steamid_cache[ply]\n if sid == nil then\n sid = ply:SteamID()\n steamid_cache[ply] = sid\n end\n return sid\nend\n\n-- hot path\nfor _, ply in ipairs(player.GetAll()) do\n local sid = GetSteamIDFast(ply)\n -- use sid repeatedly here instead of ply:SteamID()\nend\n``` \nReason: `SteamID()` is a repeated cross-boundary lookup; caching it once per player removes duplicate work in hot loops.", "durationMs": 17162, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "e8a94aa454154982", "startedAt": "2026-07-10T17:27:13.363Z", "completedAt": "2026-07-10T17:27:30.526Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14541, "completionTokens": 521, "totalTokens": 15062, "reasoningTokens": 344, "cachedTokens": 13824 }, "answerBytes": 565, "answerChars": 565 }, { "fixtureId": "gmod.perf.string-table-concat.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain 0 through 0 candidate loops.", "finalResponse": "```lua\n-- Slow: repeated string rebuilding in a loop\n-- local out = \"\"\n-- for i = 1, #pieces do\n-- out = out .. pieces[i]\n-- end\n\n-- Fast: collect pieces, join once\nlocal parts = {}\nfor i = 1, #pieces do\n parts[#parts + 1] = pieces[i]\nend\n\nlocal out = table.concat(parts)\n``` \nReason: `table.concat` joins all pieces in one pass, so it avoids the repeated reallocations and copying caused by `..` inside a loop.", "durationMs": 16354, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "b0cf9e04ae7c5dd9", "startedAt": "2026-07-10T17:27:14.426Z", "completedAt": "2026-07-10T17:27:30.782Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14540, "completionTokens": 387, "totalTokens": 14927, "reasoningTokens": 244, "cachedTokens": 13824 }, "answerBytes": 418, "answerChars": 418 }, { "fixtureId": "gmod.perf.surface-drawrect.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nsurface.SetDrawColor(r, g, b, a)\nsurface.DrawRect(x, y, w, h)\n```\nReason: `surface.DrawRect` skips the rounded-rect path and is the direct primitive for solid rectangles, so it is a bit cheaper than `draw.RoundedBox(0, ...)`.", "durationMs": 14471, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "696e8de769d9ad0e", "startedAt": "2026-07-10T17:27:15.175Z", "completedAt": "2026-07-10T17:27:29.647Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14546, "completionTokens": 338, "totalTokens": 14884, "reasoningTokens": 251, "cachedTokens": 13824 }, "answerBytes": 232, "answerChars": 232 }, { "fixtureId": "gmod.perf.surface-text.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal fontName = \"Trebuchet18\"\nlocal text = \"Hello\"\n\nhook.Add(\"HUDPaint\", \"FastSingleLineText\", function()\n surface.SetFont(fontName)\n surface.SetTextColor(255, 255, 255, 255)\n surface.SetTextPos(50, 50)\n surface.DrawText(text)\nend)\n``` \nReason: `surface.DrawText` with cached font/color/position is the lightest of these options; `draw.DrawText` and especially `SimpleTextOutlined` do extra work.", "durationMs": 16621, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "95d9fe4918ee4197", "startedAt": "2026-07-10T17:27:15.519Z", "completedAt": "2026-07-10T17:27:32.142Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14545, "completionTokens": 517, "totalTokens": 15062, "reasoningTokens": 376, "cachedTokens": 13824 }, "answerBytes": 416, "answerChars": 416 }, { "fixtureId": "gmod.perf.table-count-hash.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal len = #myTable\n```\nReason: For a sequential array-like table, the preferred Lua length operator is `#`; `table.Count` is for counting all keys in a non-sequential table.", "durationMs": 12387, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3e9bbfc8452e05e3", "startedAt": "2026-07-10T17:27:15.889Z", "completedAt": "2026-07-10T17:27:28.277Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14539, "completionTokens": 189, "totalTokens": 14728, "reasoningTokens": 126, "cachedTokens": 13824 }, "answerBytes": 182, "answerChars": 182 }, { "fixtureId": "gmod.perf.table-empty.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nt = {}\n```\nReason: `t = {}` is generally faster because it just makes `t` point at a new empty table, while `table.Empty(t)` has to clear the existing table in place; use `table.Empty(t)` only when other references must see the table emptied.", "durationMs": 14776, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3a6344977f5bd475", "startedAt": "2026-07-10T17:27:16.893Z", "completedAt": "2026-07-10T17:27:31.670Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14546, "completionTokens": 416, "totalTokens": 14962, "reasoningTokens": 335, "cachedTokens": 13824 }, "answerBytes": 249, "answerChars": 249 }, { "fixtureId": "gmod.perf.table-hasvalue.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal allowed = {\n [\"apple\"] = true,\n [\"banana\"] = true,\n [\"orange\"] = true\n}\n\nif allowed[value] then\n -- value exists in the set\nend\n```\nReason: `allowed[value]` is an O(1) hash lookup, while `table.HasValue` scans a list linearly and gets slower as the list grows.", "durationMs": 9127, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "7d064428a873fdfc", "startedAt": "2026-07-10T17:27:17.291Z", "completedAt": "2026-07-10T17:27:26.420Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14548, "completionTokens": 237, "totalTokens": 14785, "reasoningTokens": 130, "cachedTokens": 13824 }, "answerBytes": 285, "answerChars": 285 }, { "fixtureId": "gmod.perf.table-literal.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal t = {\n x = 1,\n y = 2,\n z = 3,\n}\n``` \nReason: The table constructor is faster here because Lua can create and populate the table in one step, while `t = {}` followed by assignments does extra separate writes/lookups.", "durationMs": 9215, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f95a8fc610d720ad", "startedAt": "2026-07-10T17:27:18.394Z", "completedAt": "2026-07-10T17:27:27.610Z", "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14545, "completionTokens": 243, "totalTokens": 14788, "reasoningTokens": 157, "cachedTokens": 13824 }, "answerBytes": 237, "answerChars": 237 }, { "fixtureId": "gmod.perf.table-random.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "incorrect", "detail": "Did not match a fixture answer pattern.", "finalResponse": "```lua\nlocal function FastRandom(seq)\n local n = #seq\n if n == 0 then return nil end\n return seq[math.random(n)]\nend\n```\nReason: Sequential tables are array-like, so direct indexing with `math.random(#seq)` avoids `table.Random`'s generic table traversal overhead.", "durationMs": 5540, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f3e090701953d4a9", "startedAt": "2026-07-10T17:27:21.115Z", "completedAt": "2026-07-10T17:27:26.657Z", "answerBytes": 273, "answerChars": 273, "exitCode": 0, "usage": { "promptTokens": 66, "completionTokens": 69, "totalTokens": 135, "source": "estimated" } }, { "fixtureId": "gmod.perf.x-times-x.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal y = x * x\n```\nReason: `x * x` is the fastest clear form in Lua for squaring because it avoids function-call overhead and exponentiation machinery.", "durationMs": 5204, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "79d48008f6b5174f", "startedAt": "2026-07-10T17:27:23.348Z", "completedAt": "2026-07-10T17:27:28.553Z", "answerBytes": 159, "answerChars": 159, "exitCode": 0, "usage": { "promptTokens": 58, "completionTokens": 40, "totalTokens": 98, "source": "estimated" } }, { "fixtureId": "gmod.player-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attemptIndex": 1, "status": "partial", "detail": "Uses ipairs but still allocates via player.GetAll() each time.", "finalResponse": "```lua\nfor _, ply in ipairs(player.GetAll()) do\n -- loop body\nend\n```\nReason: `player.GetAll()` returns the connected player list, and `ipairs` walks it in order.", "durationMs": 5188, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 2, "rubricVersion": "4", "promptHash": "fe3d7469535a0ca2", "startedAt": "2026-07-10T17:27:23.963Z", "completedAt": "2026-07-10T17:27:29.153Z", "answerBytes": 165, "answerChars": 165, "exitCode": 0, "usage": { "promptTokens": 80, "completionTokens": 42, "totalTokens": 122, "source": "estimated" } }, { "fixtureId": "gmod.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor _, ent in pairs(ents.FindByClass(\"prop_physics\")) do\n -- visit ent here\nend\n```\nReason: `ents.FindByClass` returns only `prop_physics` entities, so you avoid building a full map-wide entity table first.", "durationMs": 127769, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "3990c92253db85bb", "startedAt": "2026-07-10T17:27:25.258Z", "completedAt": "2026-07-10T17:29:33.029Z", "answerBytes": 216, "answerChars": 216, "exitCode": 0, "usage": { "promptTokens": 72, "completionTokens": 54, "totalTokens": 126, "source": "estimated" } }, { "fixtureId": "gmod.hook-add.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Registers a named hook with hook.Add.", "finalResponse": "```lua\nhook.Add(\"Think\", \"MyAddon.Think\", function()\nend)\n```\nReason: Registers the callback under the `Think` hook with the identifier `MyAddon.Think`, so it runs every frame.", "durationMs": 9218, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "12d2222fdb7d6b90", "startedAt": "2026-07-10T17:27:25.661Z", "completedAt": "2026-07-10T17:27:34.881Z", "answerBytes": 176, "answerChars": 176, "exitCode": 0, "usage": { "promptTokens": 78, "completionTokens": 44, "totalTokens": 122, "source": "estimated" } }, { "fixtureId": "gmod.isvalid.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Uses IsValid for entity/object validity.", "finalResponse": "```lua\nif IsValid(ent) then\n ent:SomeMethod()\nend\n```\nReason: `IsValid(ent)` is the Garry's Mod guard for nil, removed, or otherwise invalid entities before calling methods on them.", "durationMs": 8393, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "35527312e59dfc13", "startedAt": "2026-07-10T17:27:25.963Z", "completedAt": "2026-07-10T17:27:34.360Z", "answerBytes": 184, "answerChars": 184, "exitCode": 0, "usage": { "promptTokens": 74, "completionTokens": 46, "totalTokens": 120, "source": "estimated" } }, { "fixtureId": "gmod.net-receivers.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nnet.Receive(\"MyAddon.Sync\", function(len, ply)\n -- handle MyAddon.Sync here\nend)\n```\nReason: `net.Receive` registers the callback that runs when the incoming network message `MyAddon.Sync` is received.", "durationMs": 11801, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "2", "promptHash": "24d54ee8366753a6", "startedAt": "2026-07-10T17:27:26.158Z", "completedAt": "2026-07-10T17:27:37.960Z", "answerBytes": 211, "answerChars": 211, "exitCode": 0, "usage": { "promptTokens": 64, "completionTokens": 53, "totalTokens": 117, "source": "estimated" } }, { "fixtureId": "gmod.perf.angle-zero.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "partial", "detail": "Matched a partial-answer pattern.", "finalResponse": "```lua\nlocal ANGLE_ZERO = angle_zero -- read-only cached zero angle\n\n-- slow:\n-- local ang = Angle()\n\n-- fast:\nlocal ang = ANGLE_ZERO\n```\nReason: `angle_zero` is preallocated, so reusing it avoids constructing a new `Angle` userdata and the GC churn from `Angle()`.", "durationMs": 93025, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f6bc97bf98280d62", "startedAt": "2026-07-10T17:27:26.421Z", "completedAt": "2026-07-10T17:28:59.448Z", "answerBytes": 265, "answerChars": 265, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 67, "totalTokens": 138, "source": "estimated" } }, { "fixtureId": "gmod.perf.config-var.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal addon_color = Color(255, 255, 255)\nlocal color = addon_color\n```\nReason: a local/upvalue read is faster than repeated `myaddon.config.color` table lookups, and faster than a global.", "durationMs": 25042, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "bcc4ff8eb162b699", "startedAt": "2026-07-10T17:27:26.658Z", "completedAt": "2026-07-10T17:27:51.702Z", "answerBytes": 194, "answerChars": 194, "exitCode": 0, "usage": { "promptTokens": 82, "completionTokens": 49, "totalTokens": 131, "source": "estimated" } }, { "fixtureId": "gmod.perf.darkrpvar.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal jobName = ply:getDarkRPVar(\"job\")\n```\nReason: `getDarkRPVar(\"job\")` is DarkRP's public accessor for the synced job name, so it is the supported API instead of reading the raw NWString directly.", "durationMs": 46748, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "02a37daa699cded3", "startedAt": "2026-07-10T17:27:27.250Z", "completedAt": "2026-07-10T17:28:14.000Z", "answerBytes": 206, "answerChars": 206, "exitCode": 0, "usage": { "promptTokens": 66, "completionTokens": 52, "totalTokens": 118, "source": "estimated" } }, { "fixtureId": "gmod.perf.disttosqr.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nif v1:DistToSqr(v2) < threshold * threshold then\n -- within range\nend\n```\nReason: `DistToSqr` avoids the square root, so it is faster and the right choice when you only need a threshold check.", "durationMs": 6162, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "a5772470c6859cdc", "startedAt": "2026-07-10T17:27:27.611Z", "completedAt": "2026-07-10T17:27:33.774Z", "answerBytes": 202, "answerChars": 202, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 51, "totalTokens": 122, "source": "estimated" } }, { "fixtureId": "gmod.perf.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor _, ent in ents.Iterator() do\n -- visit ent\nend\n```\nReason: `ents.Iterator()` walks entities without materializing `ents.GetAll()`'s table, so it is the better choice when you only need to visit each entity once.", "durationMs": 75779, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "306d3f1d05f28246", "startedAt": "2026-07-10T17:27:27.676Z", "completedAt": "2026-07-10T17:28:43.456Z", "answerBytes": 225, "answerChars": 225, "exitCode": 0, "usage": { "promptTokens": 66, "completionTokens": 57, "totalTokens": 123, "source": "estimated" } }, { "fixtureId": "gmod.perf.find-ents-near.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal pos = ply:GetPos()\nlocal r = 512\nlocal mins = pos - Vector(r, r, r)\nlocal maxs = pos + Vector(r, r, r)\n\nlocal nearby = ents.FindInBox(mins, maxs)\n```\nReason: `FindInBox` is one of the fastest built-in broad-phase spatial queries in common GMod benchmarks, and it fits a cheap “near this player” check without the visibility semantics of `FindInPVS`. ", "durationMs": 13625, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "07c583fc3d9fe09f", "startedAt": "2026-07-10T17:27:28.278Z", "completedAt": "2026-07-10T17:27:41.905Z", "answerBytes": 367, "answerChars": 363, "exitCode": 0, "usage": { "promptTokens": 83, "completionTokens": 91, "totalTokens": 174, "source": "estimated" } }, { "fixtureId": "gmod.perf.find-players-near.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal function PlayersInRadius(pos, radius)\n local radiusSqr = radius * radius\n local out = {}\n\n for _, ply in ipairs(player.GetAll()) do\n if pos:DistToSqr(ply:GetPos()) <= radiusSqr then\n out[#out + 1] = ply\n end\n end\n\n return out\nend\n``` \nReason: `player.GetAll()` iterates only players, so it avoids scanning every entity in `ents.FindInSphere`, and `DistToSqr` skips the expensive square root.", "durationMs": 11731, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "d139ba2e8ee38a2f", "startedAt": "2026-07-10T17:27:28.554Z", "completedAt": "2026-07-10T17:27:40.286Z", "answerBytes": 444, "answerChars": 444, "exitCode": 0, "usage": { "promptTokens": 79, "completionTokens": 111, "totalTokens": 190, "source": "estimated" } }, { "fixtureId": "gmod.perf.for-vs-while.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nfor i = 1, n do\n -- work\nend\n```\nReason: a numeric `for` loop is generally faster than `while` for a tight 1..n counter loop in GMod Lua.", "durationMs": 7313, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "b3401a190a8aecf0", "startedAt": "2026-07-10T17:27:29.153Z", "completedAt": "2026-07-10T17:27:36.468Z", "answerBytes": 147, "answerChars": 147, "exitCode": 0, "usage": { "promptTokens": 65, "completionTokens": 37, "totalTokens": 102, "source": "estimated" } }, { "fixtureId": "gmod.perf.hook-once.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\n-- Faster: one hook dispatch per tick, pass the full player list.\nhook.Run(\"MyCustomTick\", player.GetAll())\n\n-- Slower: dispatches once per player.\nfor _, ply in ipairs(player.GetAll()) do\n hook.Run(\"MyCustomTickPlayer\", ply)\nend\n``` \nReason: `hook.Run` has fixed dispatch overhead, so one call per tick is cheaper than one call per player.", "durationMs": 11059, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "ed1d3e09249fc8bb", "startedAt": "2026-07-10T17:27:29.415Z", "completedAt": "2026-07-10T17:27:40.476Z", "answerBytes": 350, "answerChars": 350, "exitCode": 0, "usage": { "promptTokens": 84, "completionTokens": 88, "totalTokens": 172, "source": "estimated" } }, { "fixtureId": "gmod.perf.hudpaint-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nhook.Add(\"HUDPaint\", \"MyHUD\", function()\n local lp = LocalPlayer()\n local health = lp:Health()\n\n draw.SimpleText(\"HP: \" .. health, \"Trebuchet24\", 20, 20, color_white)\n draw.SimpleText(\"Health: \" .. health, \"Trebuchet24\", 20, 40, color_white)\n draw.RoundedBox(4, 20, 60, health * 2, 16, color_red)\nend)\n``` \nReason: Cache `LocalPlayer()` and `:Health()` once per frame instead of calling `LocalPlayer():Health()` three times.", "durationMs": 12411, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3def450f3158ceff", "startedAt": "2026-07-10T17:27:29.570Z", "completedAt": "2026-07-10T17:27:41.983Z", "answerBytes": 446, "answerChars": 446, "exitCode": 0, "usage": { "promptTokens": 76, "completionTokens": 112, "totalTokens": 188, "source": "estimated" } }, { "fixtureId": "gmod.perf.local-color.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal WHITE = color_white\n\nhook.Add(\"HUDPaint\", \"FastPaintExample\", function()\n surface.SetDrawColor(WHITE)\n surface.DrawRect(10, 10, 100, 20)\nend)\n``` \nReason: `Color(255,255,255)` allocates a new color table; reusing `color_white` or a cached local avoids that per-iteration work.", "durationMs": 12903, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "16a7bed1b931d467", "startedAt": "2026-07-10T17:27:29.648Z", "completedAt": "2026-07-10T17:27:42.553Z", "answerBytes": 295, "answerChars": 295, "exitCode": 0, "usage": { "promptTokens": 63, "completionTokens": 74, "totalTokens": 137, "source": "estimated" } }, { "fixtureId": "gmod.perf.local-vs-global.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal sin = math.sin\n\nfor i = 1, n do\n sum = sum + sin(i)\nend\n```\nReason: caching `math.sin` in a local avoids repeated global/table lookups inside the hot loop.", "durationMs": 8887, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "c116d8c651190647", "startedAt": "2026-07-10T17:27:30.527Z", "completedAt": "2026-07-10T17:27:39.415Z", "answerBytes": 171, "answerChars": 171, "exitCode": 0, "usage": { "promptTokens": 62, "completionTokens": 43, "totalTokens": 105, "source": "estimated" } }, { "fixtureId": "gmod.perf.localplayer-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal LocalPlayer = LocalPlayer\n\nhook.Add(\"HUDPaint\", \"FastLocalPlayer\", function()\n local lp = LocalPlayer()\n if not IsValid(lp) then return end\n\n -- Reuse lp instead of calling LocalPlayer() repeatedly.\n if lp:Alive() then\n -- hot-path logic here\n end\nend)\n``` \nReason: `LocalPlayer` is cached as a local upvalue, and the player entity is fetched once per hot path so the same `lp` value can be reused without repeated lookups.", "durationMs": 25613, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "7398105af47b079e", "startedAt": "2026-07-10T17:27:30.783Z", "completedAt": "2026-07-10T17:27:56.397Z", "answerBytes": 458, "answerChars": 458, "exitCode": 0, "usage": { "promptTokens": 70, "completionTokens": 115, "totalTokens": 185, "source": "estimated" } }, { "fixtureId": "gmod.perf.looking-at.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "partial", "detail": "Matched a partial-answer pattern.", "finalResponse": "```lua\nlocal LOOK_DOT = 0.70710678 -- cos(45°); raise for stricter, lower for looser\n\nlocal function LookingAtEachOther(a, b)\n if not IsValid(a) or not IsValid(b) then return false end\n\n local ab = b:EyePos() - a:EyePos()\n ab:Normalize()\n\n return a:GetAimVector():Dot(ab) >= LOOK_DOT\n and b:GetAimVector():Dot(-ab) >= LOOK_DOT\nend\n``` \nReason: Dot products measure alignment directly, so you avoid `math.acos` and keep the check to a few vector ops.", "durationMs": 61917, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "20b22f678a4b08d7", "startedAt": "2026-07-10T17:27:30.946Z", "completedAt": "2026-07-10T17:28:32.864Z", "answerBytes": 468, "answerChars": 467, "exitCode": 0, "usage": { "promptTokens": 91, "completionTokens": 117, "totalTokens": 208, "source": "estimated" } }, { "fixtureId": "gmod.perf.meta-vs-arg.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nSomeHelper(ply)\n```\nReason: `SomeHelper(ply)` is generally a little faster and clearer for a helper you control because it avoids method-style lookup/sugar; the difference is tiny, but the plain function call is the leaner choice.", "durationMs": 24924, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "de332f7356830dfa", "startedAt": "2026-07-10T17:27:31.671Z", "completedAt": "2026-07-10T17:27:56.596Z", "answerBytes": 237, "answerChars": 237, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 60, "totalTokens": 131, "source": "estimated" } }, { "fixtureId": "gmod.perf.msg-vs-print.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nMsgN(\"debug: \", value)\n```\nReason: `MsgN` is the direct console output primitive in Garry's Mod, so it avoids `print`'s extra formatting work in a hot path.", "durationMs": 46851, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "6cbb4e4b2f19dd24", "startedAt": "2026-07-10T17:27:32.143Z", "completedAt": "2026-07-10T17:28:18.996Z", "answerBytes": 163, "answerChars": 163, "exitCode": 0, "usage": { "promptTokens": 63, "completionTokens": 41, "totalTokens": 104, "source": "estimated" } }, { "fixtureId": "gmod.perf.pairs-ipairs-for.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "incorrect", "detail": "Did not match a fixture answer pattern.", "finalResponse": "```lua\nlocal sum = 0\nlocal n = #t\n\nfor i = 1, n do\n sum = sum + t[i]\nend\n```\nReason: Numeric `for` is typically fastest for a dense sequential array in Garry's Mod Lua; `ipairs` adds iterator overhead and `pairs` is the slowest and least predictable for arrays.", "durationMs": 9100, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "de5e4875fcf3e600", "startedAt": "2026-07-10T17:27:33.775Z", "completedAt": "2026-07-10T17:27:42.877Z", "answerBytes": 264, "answerChars": 264, "exitCode": 0, "usage": { "promptTokens": 73, "completionTokens": 66, "totalTokens": 139, "source": "estimated" } }, { "fixtureId": "gmod.perf.plain-lua-vs-fn.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal function fastFilterMap(src, predicate, mapper)\n local out = {}\n local n = 0\n\n for i = 1, #src do\n local v = src[i]\n if predicate(v) then\n n = n + 1\n out[n] = mapper(v)\n end\n end\n\n return out\nend\n\n-- Example: keep active players and map to their job names.\nlocal jobNames = fastFilterMap(players, function(p)\n return IsValid(p) and p:Alive()\nend, function(p)\n return p:getDarkRPVar(\"job\")\nend)\n``` \nReason: one pass, no intermediate tables, and fewer closure/combinator calls than chaining DarkRP fn filter/map helpers.", "durationMs": 11717, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "ead0a98b3ff6a388", "startedAt": "2026-07-10T17:27:34.361Z", "completedAt": "2026-07-10T17:27:46.079Z", "answerBytes": 595, "answerChars": 595, "exitCode": 0, "usage": { "promptTokens": 76, "completionTokens": 149, "totalTokens": 225, "source": "estimated" } }, { "fixtureId": "gmod.perf.player-cache-hooks.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nif SERVER then\n local cachedPlayers = player.GetAll()\n local playersDirty = false\n\n local function InvalidatePlayerCache()\n playersDirty = true\n end\n\n hook.Add(\"PlayerInitialSpawn\", \"PlayerCacheJoin\", InvalidatePlayerCache)\n hook.Add(\"PlayerDisconnected\", \"PlayerCacheLeave\", InvalidatePlayerCache)\n\n hook.Add(\"Tick\", \"UseCachedPlayers\", function()\n if playersDirty then\n cachedPlayers = player.GetAll()\n playersDirty = false\n end\n\n for i = 1, #cachedPlayers do\n local ply = cachedPlayers[i]\n -- per-player work here\n end\n end)\nend\n``` \nReason: `player.GetAll()` now runs only after a join or leave marks the cache dirty, so the per-tick hook reuses the same table instead of rebuilding it every frame.", "durationMs": 111113, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "a834817afe9e6c25", "startedAt": "2026-07-10T17:27:34.881Z", "completedAt": "2026-07-10T17:29:25.996Z", "answerBytes": 812, "answerChars": 812, "exitCode": 0, "usage": { "promptTokens": 88, "completionTokens": 203, "totalTokens": 291, "source": "estimated" } }, { "fixtureId": "gmod.perf.ply-table-index.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal PlayerTemp = setmetatable({}, { __mode = \"k\" }) -- weak keys so removed players do not leak\n\nlocal function SetTemp(ply, value)\n PlayerTemp[ply] = value\nend\n\nlocal function GetTemp(ply)\n return PlayerTemp[ply]\nend\n\n-- usage\nSetTemp(ply, {\n someField = value\n})\n\nlocal temp = GetTemp(ply)\nlocal field = temp and temp.someField\n``` \nReason: a plain upvalue table lookup is cheaper than going through `ply.someField`, and weak keys keep temporary per-player data from leaking after the player is gone.", "durationMs": 29187, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f18cc80df8999147", "startedAt": "2026-07-10T17:27:36.468Z", "completedAt": "2026-07-10T17:28:05.657Z", "answerBytes": 520, "answerChars": 520, "exitCode": 0, "usage": { "promptTokens": 69, "completionTokens": 130, "totalTokens": 199, "source": "estimated" } }, { "fixtureId": "gmod.perf.seq-tables.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\n-- Faster for iteration: sequential numeric array\nlocal values = {\n \"health\",\n \"armor\",\n \"speed\",\n \"ammo\",\n}\n\nfor i = 1, #values do\n local value = values[i]\n -- use value\nend\n``` \nReason: Sequential number-indexed arrays are faster to iterate than string-keyed maps because Lua can traverse the array part with simple indexed access instead of hashing string keys.", "durationMs": 12191, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "4778b0220c3f91a3", "startedAt": "2026-07-10T17:27:37.961Z", "completedAt": "2026-07-10T17:27:50.154Z", "answerBytes": 389, "answerChars": 389, "exitCode": 0, "usage": { "promptTokens": 79, "completionTokens": 98, "totalTokens": 177, "source": "estimated" } }, { "fixtureId": "gmod.perf.setdrawcolor-split.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nsurface.SetDrawColor(r, g, b, a)\n```\nReason: Raw numeric arguments are the faster hot-path style than passing a Color object.", "durationMs": 16405, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "d110fde8a37bbe59", "startedAt": "2026-07-10T17:27:39.416Z", "completedAt": "2026-07-10T17:27:55.822Z", "answerBytes": 132, "answerChars": 132, "exitCode": 0, "usage": { "promptTokens": 71, "completionTokens": 33, "totalTokens": 104, "source": "estimated" } }, { "fixtureId": "gmod.perf.sqlite-vs-mysql.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal rows = sql.Query(\"SELECT * FROM player_data WHERE steamid = \" .. sql.SQLStr(steamid))\n```\nReason: Local SQLite is usually lower latency for a single Garry's Mod server because it stays on the same machine and avoids the network round trip and remote connection overhead of MySQL.", "durationMs": 10599, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "4a3f5d2a20d71ac9", "startedAt": "2026-07-10T17:27:40.287Z", "completedAt": "2026-07-10T17:27:50.888Z", "answerBytes": 292, "answerChars": 292, "exitCode": 0, "usage": { "promptTokens": 100, "completionTokens": 73, "totalTokens": 173, "source": "estimated" } }, { "fixtureId": "gmod.perf.steamid-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal steamid_cache = {}\n\nlocal function SteamIDCached(ply)\n local sid = steamid_cache[ply]\n if sid == nil then\n sid = ply:SteamID()\n steamid_cache[ply] = sid\n end\n return sid\nend\n\nhook.Add(\"PlayerDisconnected\", \"SteamIDCached_Clear\", function(ply)\n steamid_cache[ply] = nil\nend)\n\n-- hot path\nlocal sid = SteamIDCached(ply)\nif sid == target_sid then\n -- ...\nend\n```\nReason: memoizes the SteamID per player so repeated hot-path checks do a table lookup instead of calling into GMod every time.", "durationMs": 40856, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "e8a94aa454154982", "startedAt": "2026-07-10T17:27:40.476Z", "completedAt": "2026-07-10T17:28:21.334Z", "answerBytes": 531, "answerChars": 531, "exitCode": 0, "usage": { "promptTokens": 63, "completionTokens": 133, "totalTokens": 196, "source": "estimated" } }, { "fixtureId": "gmod.perf.string-table-concat.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain 0 through 0 candidate loops.", "finalResponse": "```lua\n-- Slow: repeated .. concatenation copies the whole string over and over.\n-- local s = \"\"\n-- for i = 1, n do\n-- s = s .. chunks[i]\n-- end\n\nlocal parts = {}\nfor i = 1, n do\n parts[i] = chunks[i]\nend\n\nlocal s = table.concat(parts)\n``` \nReason: `table.concat` builds the final string in one pass, so it avoids the repeated reallocations and copying that make `..` in a loop slow.", "durationMs": 14132, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "b0cf9e04ae7c5dd9", "startedAt": "2026-07-10T17:27:41.905Z", "completedAt": "2026-07-10T17:27:56.039Z", "answerBytes": 386, "answerChars": 386, "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14546, "completionTokens": 652, "totalTokens": 15198, "reasoningTokens": 516, "cachedTokens": 13824 } }, { "fixtureId": "gmod.perf.surface-drawrect.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nsurface.SetDrawColor(color)\nsurface.DrawRect(x, y, w, h)\n```\nReason: `surface.DrawRect` draws a plain rectangle directly, so it avoids the extra rounded-box logic in `draw.RoundedBox(0, ...)`.", "durationMs": 10319, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "696e8de769d9ad0e", "startedAt": "2026-07-10T17:27:41.984Z", "completedAt": "2026-07-10T17:27:52.304Z", "answerBytes": 199, "answerChars": 199, "exitCode": 0, "usage": { "promptTokens": 69, "completionTokens": 50, "totalTokens": 119, "source": "estimated" } }, { "fixtureId": "gmod.perf.surface-text.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal font = \"Trebuchet24\"\nlocal text = \"Hello world\"\n\nhook.Add(\"HUDPaint\", \"FastTextExample\", function()\n surface.SetFont(font)\n surface.SetTextColor(255, 255, 255, 255)\n surface.SetTextPos(100, 100)\n surface.DrawText(text)\nend)\n```\nReason: `surface.DrawText` with the `surface.*` setup calls is the lowest-overhead path for plain single-line text; `draw.DrawText` and `SimpleTextOutlined` add extra wrapper, alignment, and outline work.", "durationMs": 32271, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "95d9fe4918ee4197", "startedAt": "2026-07-10T17:27:42.553Z", "completedAt": "2026-07-10T17:28:14.826Z", "answerBytes": 457, "answerChars": 457, "exitCode": 0, "usage": { "promptTokens": 76, "completionTokens": 115, "totalTokens": 191, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-count-hash.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal len = #myTable\n```\nReason: `#` is the standard length operator for sequential array-like tables in Lua.", "durationMs": 6624, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3e9bbfc8452e05e3", "startedAt": "2026-07-10T17:27:42.877Z", "completedAt": "2026-07-10T17:27:49.503Z", "answerBytes": 116, "answerChars": 116, "exitCode": 0, "usage": { "promptTokens": 62, "completionTokens": 29, "totalTokens": 91, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-empty.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nt = {}\n```\nReason: `t = {}` is generally faster because it drops the old table and makes a fresh empty one, while `table.Empty(t)` has to walk and clear the existing table; use `table.Empty(t)` only if other code still holds that table object.", "durationMs": 10811, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "3a6344977f5bd475", "startedAt": "2026-07-10T17:27:46.080Z", "completedAt": "2026-07-10T17:27:56.893Z", "answerBytes": 250, "answerChars": 250, "exitCode": 0, "usage": { "promptTokens": 73, "completionTokens": 63, "totalTokens": 136, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-hasvalue.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "protocol_error", "detail": "Answer must contain 0 through 0 candidate loops.", "finalResponse": "```lua\nlocal list = { \"weapon_pistol\", \"weapon_smg1\", \"weapon_ar2\" }\nlocal set = {}\n\nfor _, v in ipairs(list) do\n set[v] = true\nend\n\nif set[weaponClass] then\n -- value exists in the set\nend\n``` \nReason: direct table lookup is O(1) on average, while `table.HasValue` scans the list linearly.", "durationMs": 10660, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "7d064428a873fdfc", "startedAt": "2026-07-10T17:27:49.503Z", "completedAt": "2026-07-10T17:28:00.165Z", "answerBytes": 296, "answerChars": 296, "exitCode": 0, "usage": { "source": "provider", "promptTokens": 14546, "completionTokens": 628, "totalTokens": 15174, "reasoningTokens": 516, "cachedTokens": 13824 } }, { "fixtureId": "gmod.perf.table-literal.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal t = { x = 1, y = 2, z = 3 }\n```\nReason: The table constructor creates and fills the table in one step, while assigning fields after `{}` does extra work.", "durationMs": 23787, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f95a8fc610d720ad", "startedAt": "2026-07-10T17:27:50.154Z", "completedAt": "2026-07-10T17:28:13.943Z", "answerBytes": 166, "answerChars": 166, "exitCode": 0, "usage": { "promptTokens": 70, "completionTokens": 42, "totalTokens": 112, "source": "estimated" } }, { "fixtureId": "gmod.perf.table-random.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "incorrect", "detail": "Did not match a fixture answer pattern.", "finalResponse": "```lua\nlocal function FastRandom(t)\n local n = #t\n if n == 0 then return nil end\n return t[math.random(n)]\nend\n``` \nReason: For a sequential array, `#t` plus direct indexing avoids the extra work `table.Random` does to handle non-sequential tables.", "durationMs": 10653, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "f3e090701953d4a9", "startedAt": "2026-07-10T17:27:50.889Z", "completedAt": "2026-07-10T17:28:01.544Z", "answerBytes": 257, "answerChars": 257, "exitCode": 0, "usage": { "promptTokens": 66, "completionTokens": 65, "totalTokens": 131, "source": "estimated" } }, { "fixtureId": "gmod.perf.x-times-x.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "pass", "detail": "Matched a pass pattern.", "finalResponse": "```lua\nlocal y = x * x\n```\nReason: `x * x` avoids function-call overhead and is the fastest simple way to square a number in Lua.", "durationMs": 5905, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 1, "rubricVersion": "1", "promptHash": "79d48008f6b5174f", "startedAt": "2026-07-10T17:27:51.703Z", "completedAt": "2026-07-10T17:27:57.611Z", "answerBytes": 129, "answerChars": 129, "exitCode": 0, "usage": { "promptTokens": 58, "completionTokens": 33, "totalTokens": 91, "source": "estimated" } }, { "fixtureId": "gmod.player-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attemptIndex": 1, "status": "partial", "detail": "Uses ipairs but still allocates via player.GetAll() each time.", "finalResponse": "```lua\nfor _, ply in ipairs(player.GetAll()) do\n -- visit ply here\nend\n```\nReason: `player.GetAll()` returns an array-like list of connected players, so `ipairs` visits each one in order.", "durationMs": 7160, "version": "codex-cli 0.144.0-alpha.4", "fixtureVersion": 2, "rubricVersion": "4", "promptHash": "fe3d7469535a0ca2", "startedAt": "2026-07-10T17:27:52.305Z", "completedAt": "2026-07-10T17:27:59.466Z", "answerBytes": 190, "answerChars": 190, "exitCode": 0, "usage": { "promptTokens": 80, "completionTokens": 48, "totalTokens": 128, "source": "estimated" } } ], "summary": { "statusCounts": { "pass": 251, "partial": 19, "incorrect": 15, "protocol_error": 19, "policy_violation": 0, "timeout": 0, "unavailable": 0, "unsupported": 0, "trace_error": 0, "scored": 285 }, "groups": [ { "fixtureId": "gmod.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.hook-add.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.hook-add.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.hook-add.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.hook-add.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.hook-add.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.hook-add.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.hook-add.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.hook-add.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.isvalid.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.isvalid.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.isvalid.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.isvalid.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.isvalid.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.isvalid.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.isvalid.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.isvalid.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.net-receivers.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.net-receivers.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.net-receivers.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.net-receivers.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.net-receivers.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.net-receivers.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.net-receivers.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.net-receivers.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.angle-zero.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "incorrect", "meanScore": 0, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.angle-zero.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "incorrect", "meanScore": 0, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.angle-zero.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.angle-zero.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.angle-zero.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "incorrect", "meanScore": 0, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.angle-zero.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "incorrect", "meanScore": 0, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.angle-zero.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "incorrect", "meanScore": 0, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.angle-zero.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.config-var.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.config-var.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.config-var.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.config-var.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.config-var.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.config-var.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.config-var.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.config-var.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "incorrect", "meanScore": 0, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.darkrpvar.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.darkrpvar.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.darkrpvar.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.darkrpvar.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.darkrpvar.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.darkrpvar.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.darkrpvar.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.darkrpvar.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.disttosqr.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.disttosqr.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.disttosqr.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.disttosqr.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.disttosqr.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.disttosqr.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.disttosqr.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.disttosqr.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.ents-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.find-ents-near.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.find-ents-near.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.find-ents-near.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.find-ents-near.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.find-ents-near.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.find-ents-near.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.find-ents-near.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.find-ents-near.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.find-players-near.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.find-players-near.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.find-players-near.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.find-players-near.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.find-players-near.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.find-players-near.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.find-players-near.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.find-players-near.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.for-vs-while.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.for-vs-while.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.for-vs-while.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.for-vs-while.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.for-vs-while.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.for-vs-while.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.for-vs-while.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.for-vs-while.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.hook-once.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.hook-once.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.hook-once.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.hook-once.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.hook-once.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.hook-once.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.hook-once.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.hook-once.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.hudpaint-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.hudpaint-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.hudpaint-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.hudpaint-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.hudpaint-cache.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.hudpaint-cache.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.hudpaint-cache.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.hudpaint-cache.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.local-color.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.local-color.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.local-color.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.local-color.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.local-color.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.local-color.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.local-color.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.local-color.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.local-vs-global.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.local-vs-global.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.local-vs-global.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.local-vs-global.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.local-vs-global.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.local-vs-global.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.local-vs-global.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.local-vs-global.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.localplayer-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.localplayer-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.localplayer-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.localplayer-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.localplayer-cache.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.localplayer-cache.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.localplayer-cache.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.localplayer-cache.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.looking-at.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.looking-at.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.looking-at.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.looking-at.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.looking-at.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.looking-at.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.looking-at.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.looking-at.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.meta-vs-arg.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.meta-vs-arg.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.meta-vs-arg.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.meta-vs-arg.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.meta-vs-arg.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.meta-vs-arg.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.meta-vs-arg.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.meta-vs-arg.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.msg-vs-print.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.msg-vs-print.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.msg-vs-print.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.msg-vs-print.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.msg-vs-print.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.msg-vs-print.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.msg-vs-print.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.msg-vs-print.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.pairs-ipairs-for.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.pairs-ipairs-for.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.pairs-ipairs-for.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.pairs-ipairs-for.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "incorrect", "meanScore": 0, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.pairs-ipairs-for.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.pairs-ipairs-for.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.pairs-ipairs-for.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.pairs-ipairs-for.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.plain-lua-vs-fn.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.plain-lua-vs-fn.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.plain-lua-vs-fn.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.plain-lua-vs-fn.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.plain-lua-vs-fn.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.plain-lua-vs-fn.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.plain-lua-vs-fn.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.plain-lua-vs-fn.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.player-cache-hooks.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.player-cache-hooks.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.player-cache-hooks.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.player-cache-hooks.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.player-cache-hooks.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.player-cache-hooks.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.player-cache-hooks.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.player-cache-hooks.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.ply-table-index.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.ply-table-index.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.ply-table-index.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.ply-table-index.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.ply-table-index.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "incorrect", "meanScore": 0, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.ply-table-index.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.ply-table-index.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "incorrect", "meanScore": 0, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.ply-table-index.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "incorrect", "meanScore": 0, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.seq-tables.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.seq-tables.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.seq-tables.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.seq-tables.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.seq-tables.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.seq-tables.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.seq-tables.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.seq-tables.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.setdrawcolor-split.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.setdrawcolor-split.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.setdrawcolor-split.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.setdrawcolor-split.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.setdrawcolor-split.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.setdrawcolor-split.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.setdrawcolor-split.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.setdrawcolor-split.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.sqlite-vs-mysql.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.sqlite-vs-mysql.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.sqlite-vs-mysql.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.sqlite-vs-mysql.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.sqlite-vs-mysql.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.sqlite-vs-mysql.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.sqlite-vs-mysql.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.sqlite-vs-mysql.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.steamid-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.steamid-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.steamid-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.steamid-cache.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.steamid-cache.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.steamid-cache.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.steamid-cache.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.steamid-cache.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.string-table-concat.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.string-table-concat.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.string-table-concat.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.string-table-concat.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.string-table-concat.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.string-table-concat.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.string-table-concat.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.string-table-concat.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.surface-drawrect.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.surface-drawrect.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.surface-drawrect.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.surface-drawrect.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.surface-drawrect.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.surface-drawrect.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.surface-drawrect.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.surface-drawrect.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.surface-text.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.surface-text.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.surface-text.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.surface-text.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.surface-text.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.surface-text.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.surface-text.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.surface-text.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-count-hash.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-count-hash.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-count-hash.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-count-hash.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-count-hash.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-count-hash.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-count-hash.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-count-hash.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-empty.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-empty.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-empty.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-empty.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-empty.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-empty.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-empty.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-empty.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-hasvalue.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-hasvalue.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-hasvalue.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-hasvalue.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "protocol_error", "meanScore": null, "scoredAttempts": 0 }, { "fixtureId": "gmod.perf.table-hasvalue.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-hasvalue.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "incorrect", "meanScore": 0, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-hasvalue.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-hasvalue.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-literal.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-literal.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-literal.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-literal.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-literal.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-literal.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-literal.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-literal.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-random.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "incorrect", "meanScore": 0, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-random.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-random.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-random.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "incorrect", "meanScore": 0, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-random.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-random.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-random.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.table-random.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "incorrect", "meanScore": 0, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.x-times-x.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.x-times-x.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.x-times-x.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.x-times-x.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.x-times-x.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.x-times-x.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.x-times-x.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.perf.x-times-x.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.player-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@high", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.player-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@low", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "incorrect", "meanScore": 0, "scoredAttempts": 1 }, { "fixtureId": "gmod.player-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@medium", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.player-iterator.v1", "adapterId": "codex", "model": "gpt-5.4-mini@xhigh", "attempts": 1, "passCount": 0, "passAtK": false, "bestStatus": "partial", "meanScore": 0.5, "scoredAttempts": 1 }, { "fixtureId": "gmod.player-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@high", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.player-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@low", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.player-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@medium", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 }, { "fixtureId": "gmod.player-iterator.v1", "adapterId": "codex", "model": "gpt-5.4@xhigh", "attempts": 1, "passCount": 1, "passAtK": true, "bestStatus": "pass", "meanScore": 1, "scoredAttempts": 1 } ], "passAtKRate": "251/285", "overallMeanScore": 0.9140350877192982 }, "metadata": { "schemaVersion": 3, "runId": "6d5384a1-ae22-496a-83dd-92c789967f27", "startedAt": "2026-07-10T17:24:54.308Z", "completedAt": "2026-07-10T17:29:33.033Z", "repeat": 1, "concurrency": 20, "fixtureIds": [ "gmod.ents-iterator.v1", "gmod.hook-add.v1", "gmod.isvalid.v1", "gmod.net-receivers.v1", "gmod.perf.angle-zero.v1", "gmod.perf.config-var.v1", "gmod.perf.darkrpvar.v1", "gmod.perf.disttosqr.v1", "gmod.perf.ents-iterator.v1", "gmod.perf.find-ents-near.v1", "gmod.perf.find-players-near.v1", "gmod.perf.for-vs-while.v1", "gmod.perf.hook-once.v1", "gmod.perf.hudpaint-cache.v1", "gmod.perf.local-color.v1", "gmod.perf.local-vs-global.v1", "gmod.perf.localplayer-cache.v1", "gmod.perf.looking-at.v1", "gmod.perf.meta-vs-arg.v1", "gmod.perf.msg-vs-print.v1", "gmod.perf.pairs-ipairs-for.v1", "gmod.perf.plain-lua-vs-fn.v1", "gmod.perf.player-cache-hooks.v1", "gmod.perf.ply-table-index.v1", "gmod.perf.seq-tables.v1", "gmod.perf.setdrawcolor-split.v1", "gmod.perf.sqlite-vs-mysql.v1", "gmod.perf.steamid-cache.v1", "gmod.perf.string-table-concat.v1", "gmod.perf.surface-drawrect.v1", "gmod.perf.surface-text.v1", "gmod.perf.table-count-hash.v1", "gmod.perf.table-empty.v1", "gmod.perf.table-hasvalue.v1", "gmod.perf.table-literal.v1", "gmod.perf.table-random.v1", "gmod.perf.x-times-x.v1", "gmod.player-iterator.v1" ], "fixtureCount": 38, "attemptCount": 304, "adapters": [ "codex" ], "models": [ "codex/gpt-5.4-mini@high", "codex/gpt-5.4-mini@low", "codex/gpt-5.4-mini@medium", "codex/gpt-5.4-mini@xhigh", "codex/gpt-5.4@high", "codex/gpt-5.4@low", "codex/gpt-5.4@medium", "codex/gpt-5.4@xhigh" ], "keepRaw": true, "durationSeconds": 279, "usageTotals": { "promptTokens": 1662565, "completionTokens": 59193, "reasoningTokens": 31334, "totalTokens": 1721758, "cachedTokens": 1170944, "cacheWriteTokens": 0, "cost": 0, "upstreamInferenceCost": 0, "attemptsWithUsage": 304, "providerUsageAttempts": 112, "estimatedUsageAttempts": 192, "avgDurationMs": 13472, "minDurationMs": 2983, "maxDurationMs": 127769, "totalDurationMs": 4095432, "totalAnswerBytes": 99339 } } }