GPT-4.1 ranks first on 5 of the 188 benchmarks with five or more models scored.
Closed weights Flagship
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MMBench-CN v1.1 | 86.2% | #9 of 178 | Official leaderboard | |
| ECLeKTic | 34.2% | #5 of 35 | Official leaderboard | |
| Vectara Hallucination Leaderboard (HHEM) | 2.00% | #24 of 148 | Official leaderboard | |
| BuzzBench | 67.1 | #4 of 20 | Official leaderboard | |
| Vectara Hallucination Leaderboard (2025 dataset) | 5.60% | #21 of 104 | Official leaderboard | |
| AraGen | 74.5% | #9 of 42 | Official leaderboard | |
| FaithJudge | 11.9% | #8 of 37 | Official leaderboard | |
| MMLU-Pro | 81.8% | #48 of 212 | Official leaderboard | |
| HELM Arabic | 81.4 | #7 of 28 | Official leaderboard | |
| SimpleQA | 41.6% | #5 of 16 | Official leaderboard | |
| RULERv2 | 89.2% | #1 of 3 | Official leaderboard | |
| LMArena (Elo) | 1,415 | #114 of 335 | Official leaderboard | |
| RewardBench 2 | 72.3% | #15 of 42 | Official leaderboard | |
| OpenAI MRCR | 57.2% | #4 of 10 | Lab model card | |
| FACTS Parametric | 51.5% | #19 of 45 | Official leaderboard | |
| Judgemark v2.1 | 76.3 | #18 of 42 | Official leaderboard | |
| IFEval | 87.4% | #61 of 135 | Lab model card | |
| Wer wird Millionär? LLM Benchmark | 7022.0 | #36 of 77 | Official leaderboard | |
| Phare | 67.1% | #29 of 62 | Official leaderboard | |
| Arena-Hard v2.0 | 50% | #11 of 23 | Official leaderboard | |
| HUMAINE | 29 | #25 of 50 | Official leaderboard | |
| Visual Aesthetic Benchmark | 29.5% | #9 of 18 | Official leaderboard | |
| Internal API instruction following (OpenAI) | 49.1% | #4 of 8 | Official leaderboard | |
| Fiction.LiveBench | 63.9% | #29 of 57 | Fiction.live leaderboard, via Epoch AI | |
| EQ-Bench 3 | 1,257 | #41 of 78 | Official leaderboard | |
| WritingBench | 70.9 | #25 of 45 | Official leaderboard | |
| Creative Writing v3 | 65.8 | #68 of 120 | Official leaderboard | |
| DiscoX | 49.6 | #4 of 7 | Official leaderboard | |
| Structured Output Benchmark | 85% | #11 of 19 | Official leaderboard | |
| Global MMLU Lite | 87.5% | #25 of 43 | Official leaderboard | |
| FACTS Grounding | 74.5% | #23 of 34 | Official leaderboard | |
| IndicGenBench | 43.7 | #9 of 13 | Official leaderboard | |
| SnorkelSpatial | 14.6% | #19 of 27 | Official leaderboard | |
| FACTS Search | 64.6% | #22 of 31 | Official leaderboard | |
| MultiNRC | 21.2% | #28 of 38 | Official leaderboard | |
| SimpleQA Verified | 31.1% | #70 of 94 | Epoch AI run | |
| BullshitBench | 14.6% | #81 of 107 | Official leaderboard | |
| GraphWalks | 61.7% | #21 of 27 | Lab model card | |
| Meeseeks Utility Rate | 69.4% | #8 of 10 | Official leaderboard | |
| BullshitBench V2 | 14% | #105 of 129 | Official leaderboard | |
| AMemGym | 39.5% | #9 of 11 | Official leaderboard | |
| MultiChallenge | 39.4% | #36 of 42 | Official leaderboard | |
| Humanity's Last Exam | 5.40% | #41 of 47 | Official leaderboard | |
| Seal-0 | 0% | #10 of 11 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MATH Level 5 | 83% | #25 of 94 | Epoch AI run | |
| SGI-Bench | 31.4% | #4 of 15 | Official leaderboard | |
| GPQA Main | 65.9% | #17 of 60 | Official leaderboard | |
| SnorkelGraph | 63% | #10 of 24 | Official leaderboard | |
| SciArena | 1,019 | #15 of 34 | Official leaderboard | |
| CORE-Bench | 33.3% | #8 of 17 | Official leaderboard | |
| ForecastBench | 0.13 | #39 of 82 | Official leaderboard | |
| Misguided Attention | 44.2% | #23 of 45 | Official leaderboard | |
| GPQA Diamond | 66.9% | #124 of 229 | Epoch AI run | |
| FutureX | 13.7 | #4 of 7 | Official leaderboard | |
| LiveMathematicianBench | 25.8% | #10 of 17 | Official leaderboard | |
| DTBench | 68.3% | #117 of 180 | Official leaderboard | |
| OTIS Mock AIME 2024-2025 | 38.3% | #140 of 212 | Epoch AI run | |
| ScholarCatalyst | 37% | #2 of 3 | Official leaderboard | |
| IneqMath | 2.50% | #23 of 34 | Official leaderboard | |
| PolyMath | 26.4% | #26 of 38 | Official leaderboard | |
| GSM8K Platinum Initial 268 | 98.5% | #13 of 19 | Official leaderboard | |
| FrontierMath | 5.52% | #39 of 55 | Official leaderboard | |
| MATH-500 | 87.2% | #28 of 38 | Official leaderboard | |
| SimpleBench | 27% | #89 of 106 | Official leaderboard | |
| AstaBench | 30.4% | #11 of 13 | Official leaderboard | |
| SnorkelSequences | 36.8% | #22 of 25 | Official leaderboard | |
| FrontierMath Tier 4 | 0% | #49 of 55 | Epoch AI run | |
| PHYBench | 13.2% | #25 of 28 | Official leaderboard | |
| ARC-AGI-2 | 0.42% | #89 of 98 | Official leaderboard | |
| ARC-AGI | 5.50% | #89 of 95 | Official leaderboard | |
| AMO-Bench | 4.10% | #20 of 21 | Official leaderboard | |
| Verifying the Verifiers (Project APE CRED benchmark) | 14% | #22 of 23 | Official leaderboard | |
| FrontierMath Tiers 1-3 v2 | 5.96% | #83 of 86 | Official leaderboard | |
| CMT-Benchmark | 4.00% | #11 of 11 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BigCodeBench-Hard | 32.8% | #8 of 96 | Official leaderboard | |
| GitTaskBench | 42.6% | #2 of 10 | Official leaderboard | |
| ArtifactsBench (Gemini 2.5 Pro 0506 judge) | 48.2% | #4 of 18 | Official leaderboard | |
| AutoCodeBench | 48% | #9 of 34 | Official leaderboard | |
| AetherCode | 10.5% | #5 of 9 | Official leaderboard | |
| SWE-bench Multimodal | 31.1% | #5 of 9 | Official leaderboard | |
| Aider Polyglot | 52.9% | #19 of 30 | Lab model card | |
| WeirdML | 39% | #82 of 112 | Official leaderboard | |
| ArtifactsBench | 46% | #15 of 20 | Official leaderboard | |
| DesignArena Game Dev | 1,070 | #109 of 143 | Official leaderboard | |
| SWE-bench-Live | 16.3% | #10 of 12 | Official leaderboard | |
| DesignArena Website | 1,032 | #126 of 151 | Official leaderboard | |
| DesignArena Code | 1,016 | #124 of 143 | Official leaderboard | |
| DesignArena UI Component | 991 | #121 of 139 | Official leaderboard | |
| ALE-Bench | 985 | #112 of 126 | Official leaderboard | |
| SWE-bench Verified | 48.5% | #31 of 32 | Epoch AI run | |
| DesignArena 3D Design | 833 | #135 of 136 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Galileo Agent Leaderboard v2 | 62% | #1 of 21 | Official leaderboard | |
| TextQuests | 11.4 | #2 of 28 | Official leaderboard | |
| Berkeley Function Calling Leaderboard (BFCL) | 68.9% | #7 of 67 | Official leaderboard | |
| GAIA | 83.1% | #8 of 64 | Official leaderboard | |
| SnakeBench | 32 | #26 of 188 | Official leaderboard | |
| Galileo Agent Leaderboard | 0.92 | #5 of 29 | Official leaderboard | |
| BFCL V4 | 54% | #16 of 65 | Official leaderboard | |
| C3-Bench | 8.98% | #10 of 30 | Official leaderboard | |
| LiveResearchBench | 65.6 | #3 of 7 | Official leaderboard | |
| MM-BrowseComp | 4.75% | #6 of 11 | Official leaderboard | |
| AppWorld Test-Challenge | 57.6% | #7 of 12 | Official leaderboard | |
| Pencil Puzzle Bench | 0% | #33 of 53 | Official leaderboard | |
| LLM Chess | -111 | #78 of 123 | Official leaderboard | |
| lmgame-Bench | 20.4 | #15 of 23 | Official leaderboard | |
| AppWorld Test-Normal | 73.2% | #8 of 12 | Official leaderboard | |
| LisanBench | 172.0 | #96 of 133 | Official leaderboard | |
| MCP-Universe | 19.9% | #24 of 30 | Official leaderboard | |
| CompileBench | 60% | #17 of 20 | Official leaderboard | |
| MCPMark | 8.07% | #24 of 28 | Official leaderboard | |
| BrowseComp-Plus | 35.4% | #6 of 7 | Official leaderboard | |
| SnorkelWordle | 62% | #17 of 19 | Official leaderboard | |
| tau2-bench | 54.7% | #9 of 10 | Official leaderboard | |
| Chess Puzzles | 11% | #84 of 93 | Epoch AI run | |
| UniClawBench | 15.3% | #10 of 10 | Official leaderboard | |
| WebHarbor | 26.4% | #4 of 4 | Official leaderboard | |
| Wiki Live Challenge | 20.7 | #2 of 2 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CaseLaw v2 | 69.9% | #3 of 53 | Official leaderboard | |
| LEXam | 57.5% | #5 of 30 | Official leaderboard | |
| TaxEval v2 | 75.1% | #24 of 129 | Official leaderboard | |
| AssetOpsBench | 65% | #1 of 3 | Official leaderboard | |
| CorpFin v2 | 63% | #47 of 119 | Official leaderboard | |
| MortgageTax | 65.9% | #35 of 87 | Official leaderboard | |
| LegalBench | 83.1% | #59 of 138 | Official leaderboard | |
| HealthBench | 47.9% | #4 of 8 | Official leaderboard | |
| HealthBench Hard | 16% | #4 of 8 | Official leaderboard | |
| SnorkelUnderwrite | 70.6% | #14 of 27 | Official leaderboard | |
| Finance Reasoning (Snorkel) | 44.3% | #11 of 21 | Official leaderboard | |
| ProfBench | 46.3% | #40 of 72 | Official leaderboard | |
| CadEval | 42% | #8 of 13 | Official leaderboard | |
| Vals Finance Agent v1 | 24.6% | #29 of 47 | Official leaderboard | |
| FinQA (Snorkel OpenEnv environment) | 62.7% | #16 of 22 | Official leaderboard | |
| SnorkelFinance | 62.7% | #16 of 22 | Official leaderboard | |
| PRBench | 36.5% | #44 of 48 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SEED-Bench-2-Plus | 73.1% | #2 of 136 | Official leaderboard | |
| A-Bench | 79.6% | #2 of 116 | Official leaderboard | |
| MM-Vet | 78.8% | #5 of 164 | Official leaderboard | |
| MMBench-EN v1.1 | 86.9% | #7 of 178 | Official leaderboard | |
| MMBench | 86.9% | #9 of 155 | Official leaderboard | |
| MMStar | 69.8% | #11 of 173 | Official leaderboard | |
| CrossVid | 45.2% | #1 of 13 | Official leaderboard | |
| CharXiv (DQ) | 87.9% | #6 of 67 | Official leaderboard | |
| VideoEval-Pro | 40.8% | #1 of 10 | Official leaderboard | |
| We-Math | 55.5% | #3 of 21 | Official leaderboard | |
| CCBench (MMBench Chinese Culture) | 73.5% | #23 of 152 | Official leaderboard | |
| OST-Bench | 53.4% | #1 of 6 | Official leaderboard | |
| MMSI-Bench | 30.9% | #6 of 28 | Official leaderboard | |
| VHELM | 69.5 | #9 of 41 | Official leaderboard | |
| MM-HELIX | 43.3% | #4 of 17 | Official leaderboard | |
| MME-CC | 32.1% | #3 of 12 | Official leaderboard | |
| CompassArena Multimodal | 1,096 | #3 of 8 | Official leaderboard | |
| MMMU | 74.8% | #45 of 103 | Lab model card | |
| VISTA | 45.3% | #26 of 54 | Official leaderboard | |
| MathCanvas-Bench | 19% | #8 of 16 | Official leaderboard | |
| Vision Arena | 1,214 | #71 of 136 | Official leaderboard | |
| MMLongBench-Doc | 49.7% | #9 of 17 | Official leaderboard | |
| GeoBench | 72% | #26 of 48 | Official leaderboard | |
| vCX Benchmark | 84.6% | #24 of 42 | Official leaderboard | |
| MathVista | 72.2% | #42 of 70 | Lab model card | |
| MADQA | 60% | #11 of 18 | Official leaderboard | |
| Math-VR | 26% | #16 of 26 | Official leaderboard | |
| Point-Bench | 33.3% | #12 of 17 | Official leaderboard | |
| VisFactor | 21.3% | #20 of 26 | Official leaderboard | |
| ZeroBench | 3.00% | #59 of 76 | Official leaderboard | |
| Video-MME | 72% | #43 of 51 | Lab model card | |
| MineBench | 1,033 | #67 of 75 | Official leaderboard | |
| CharXiv | 56.7% | #68 of 75 | Lab model card | |
| VLMsAreBiased | 16% | #5 of 5 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BioTIER Permit | 99.9% | #1 of 69 | Official leaderboard | |
| SimpleSafetyTests | 100% | #10 of 75 | Official leaderboard | |
| HELM Safety | 96.3 | #17 of 75 | Official leaderboard | |
| BaxBench | 40.8% | #9 of 34 | Official leaderboard | |
| Anthropic Red Team | 99.3% | #28 of 75 | Official leaderboard | |
| nocot-bench (NCRI) | 75.1 | #75 of 185 | Official leaderboard | |
| NYU CTF Bench | 13.5% | #3 of 7 | Official leaderboard | |
| AIR-Bench 2024 | 64.8% | #43 of 82 | Official leaderboard | |
| BABE | 36.9% | #4 of 6 | Official leaderboard | |
| KORA | 26.2% | #29 of 42 | Official leaderboard | |
| Molecular Biology Capabilities Test | 42.5% | #66 of 90 | Official leaderboard | |
| EvalDetectBench | 0.78 | #8 of 10 | Official leaderboard | |
| World-Class Bio | 31.8% | #60 of 74 | Official leaderboard | |
| CyberGym | 9.36% | #28 of 34 | Official leaderboard | |
| FORTRESS | 53% | #61 of 73 | Official leaderboard | |
| Human Pathogen Capabilities Test v2 (HPCT v2) | 34.2% | #60 of 71 | Official leaderboard | |
| Human Pathogen Capabilities Test | 38.7% | #73 of 86 | Official leaderboard | |
| BioTIER Refuse | 16.7% | #62 of 69 | Official leaderboard | |
| Lakera B3 (Backbone Breaker Benchmark) | 37.3% | #35 of 35 | Official leaderboard | |
| SUPERChem | 12.2% | #6 of 6 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| OpenVLM Leaderboard | 75.9 | #9 of 173 | Official leaderboard | |
| HELM Capabilities | 72.7 | #16 of 60 | Official leaderboard | |
| ACCoRD | 56.0 | #103 of 152 | Official leaderboard | |
| LMCA | 25.6% | #101 of 147 | Official leaderboard | |
| Conceptual Reasoning Index | 36.0 | #108 of 151 | Official leaderboard |
The Self-Reported Index fits one score to the 10 capability results OpenAI reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 130.1 to 135.0. How we compute it.
Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.
| Benchmark | Lab says | Board says | Gap | What differs |
|---|---|---|---|---|
| SWE-bench Verified | 54.6% SWE-bench Verified (477/500) | 39.6% Leaderboard, mini-SWE-agent | +15 pts | The lab evaluates 477 tasks, while the board reports a separate mini-SWE-agent run. |
| IFEval | 87.4% IFEval | 83.8% Leaderboard, strict, share of instructions followed | +3.6 pts | |
| MultiChallenge | 38.3% MultiChallenge | 39.4% Leaderboard | −1.1 pts | |
| Aider Polyglot | 52.9% Aider's polyglot: diff | 52.4% Leaderboard, aider, diff edits | +0.5 pts | |
| CharXiv | 56.7% CharXiv-R | 56.7% Leaderboard | Same |
OpenAI's models in color, against every other model on the Epoch Capabilities Index, by release date.
Source: Epoch Capabilities Index, Epoch AI (CC BY 4.0)
The benchmarks the lab chose to headline, with the scores it gave.