Acropolis

GPT-5.3 Codex

OpenAIGrade A. Evidence grade A of A to D: strongest.ProprietaryReleased 5 Feb 2026Consistent across benchmarks
Compare with peers

Domain profile

Median within-board percentile by capability domain across boards with sufficient peer overlap.

  • Agentic (4 boards): 73 (scale 0 to 100)Agentic (4 boards)73
  • Coding (4 boards): 68 (scale 0 to 100)Coding (4 boards)68
  • Finance (1 boards): 64 (scale 0 to 100)Finance (1 boards)64

Benchmark appearances

Best verified run per board, with confidence intervals and Pareto frontier status.

BenchmarkRankScoreCostStatusObserved
METR time horizon

METR time horizons · triframe_inspect/triframe_agent

#5 / 23Tied #1
349.5 min[194.9 min to 816.4 min]UnknownStandardNot stated
Terminal Bench

Epoch AI Benchmarking Hub · SageAgent

#9 / 202
78.4%[74.1% to 82.7%]UnknownStandard5 Feb 2026
Terminal Bench

Epoch AI Benchmarking Hub · Droid

#12 / 202
77.3%[73.1% to 81.5%]UnknownStandard5 Feb 2026
Terminal Bench

Epoch AI Benchmarking Hub · CodeBrain-1.5

#14 / 202
75.8%[71.9% to 79.8%]UnknownStandard5 Feb 2026
Terminal Bench

Epoch AI Benchmarking Hub · Codelia

#15 / 202
75.7%[71.5% to 80%]UnknownStandard5 Feb 2026
SWE-Bench verified

Epoch AI Benchmarking Hub · high

#16 / 35
74.8%[70.9% to 78.7%]UnknownStandard25 Feb 2026
Terminal Bench

Epoch AI Benchmarking Hub · Simple Codex

#17 / 202
75.1%[70.4% to 79.7%]UnknownStandard5 Feb 2026
Terminal Bench

Epoch AI Benchmarking Hub · Mux

#20 / 202
74.6%[69.7% to 79.5%]UnknownStandard5 Feb 2026
Terminal Bench

Epoch AI Benchmarking Hub · spoox-o-m

#23 / 202
71.5%[66.6% to 76.3%]UnknownStandard5 Feb 2026
Terminal Bench

Epoch AI Benchmarking Hub · CodeBrain-1

#24 / 202
70.3%[65.2% to 75.4%]UnknownStandard10 Feb 2026
Vending-Bench 2

Epoch AI Benchmarking Hub · Default

#25 / 72
$5,940.12UnknownStandardNot stated
Terminal Bench

Epoch AI Benchmarking Hub · IndusAGI Coding Agent

#28 / 202
69.1%[64.7% to 73.6%]UnknownStandard5 Feb 2026
Terminal Bench

Epoch AI Benchmarking Hub · Terminus 2

#36 / 202
64.7%[59.4% to 70%]UnknownStandard5 Feb 2026
LMArena WebDev

LMArena · Image to webdev · Codex CLI

#51 / 64
1,444[1,429 to 1,458]UnknownStandard15 Nov 2026
LMArena WebDev

LMArena · Webdev HTML · Codex CLI

#98 / 155
1,421[1,390 to 1,452]UnknownStandard15 Nov 2026
LMArena WebDev

LMArena · Webdev react · Codex CLI

#99 / 140
1,394[1,379 to 1,409]UnknownStandard15 Nov 2026
LMArena WebDev

LMArena · Overall · Codex CLI

#101 / 156
1,406[1,392 to 1,420]UnknownStandard15 Nov 2026
LMArena WebDev

LMArena · Webdev · Codex CLI

#101 / 156
1,406[1,392 to 1,420]UnknownStandard15 Nov 2026
LMArena WebDev

LMArena · Webdev HTML · Codex CLI

#101 / 155
1,418[1,381 to 1,455]UnknownStandard15 Nov 2026
LMArena WebDev

LMArena · Webdev react · Codex CLI

#114 / 140
1,356[1,344 to 1,368]UnknownStandard15 Nov 2026
LMArena WebDev

LMArena · Overall · Codex CLI

#116 / 156
1,371[1,360 to 1,382]UnknownStandard15 Nov 2026
LMArena WebDev

LMArena · Webdev · Codex CLI

#116 / 156
1,371[1,360 to 1,382]UnknownStandard15 Nov 2026

Closest head-to-head rivals

Direct pairwise win rates for GPT-5.3 Codex across shared benchmark boards.