GPT-5.1
GPT-5.1 ranks first on MMMU.
112
Benchmarks with a score
1
Ranked first, out of five or more models
9
In the top three
OpenAI Nov 2025
GPT-5.1
Closed weights Flagship
149.9
#64 of 271
Epoch Capabilities Index
- Humanity's Last Exam 23.7% #18 /47
- SWE-bench Verified 68% #25 /32
- Terminal-Bench 2.0 47.6% #18 /43
Where it ranks
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.
Knowledge and language
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SuperGPQA | 66.3% | #3 of 58 | Official leaderboard | |
| CL-bench | 23.7% | #2 of 24 | Official leaderboard | |
| MMLU-Pro | 86.4% | #22 of 212 | Official leaderboard | |
| LMArena (Elo) | 1,456 | #54 of 333 | Official leaderboard | |
| HELM Arabic | 81.7 | #5 of 28 | Official leaderboard | |
| CL-bench Life | 17.3% | #3 of 16 | Official leaderboard | |
| Wer wird Millionär? LLM Benchmark | 114156.0 | #15 of 76 | Official leaderboard | |
| HELM Arabic Enterprise | 73.1 | #5 of 23 | Official leaderboard | |
| Phare | 72.8% | #14 of 62 | Official leaderboard | |
| EQ-Bench 3 | 1,527 | #18 of 78 | Official leaderboard | |
| FACTS Multimodal | 41.8% | #11 of 39 | Official leaderboard | |
| MultiNRC | 49% | #11 of 38 | Official leaderboard | |
| SkateBench v2 | 71.8% | #7 of 21 | Official leaderboard | |
| BrowseComp Long Context 128k | 90% | #1 of 3 | Model card | |
| MultiChallenge | 63.4% | #12 of 35 | Official leaderboard | |
| Humanity's Last Exam | 23.7% | #18 of 47 | Official leaderboard | |
| SimpleQA Verified | 34.9% | #26 of 56 | Official leaderboard | |
| BullshitBench | 25.4% | #57 of 107 | Official leaderboard | |
| FACTS Parametric | 43.2% | #24 of 45 | Official leaderboard | |
| FACTS Benchmark Suite | 41.8% | #22 of 40 | Official leaderboard | |
| BullshitBench V2 | 25% | #77 of 122 | Official leaderboard | |
| Vectara Hallucination Leaderboard (2025 dataset) | 10.9% | #71 of 104 | Official leaderboard | |
| FACTS Search | 62.4% | #25 of 31 | Official leaderboard | |
| Document Arena | 1,416 | #36 of 37 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CMIMC 2025 | 94.4% | #4 of 33 | Official leaderboard | |
| SMT 2025 | 91% | #7 of 40 | Official leaderboard | |
| HMMT 2025 | 93.3% | #10 of 53 | Official leaderboard | |
| AIME 2025 | 94% | #27 of 143 | Model card | |
| General365 | 58.2% | #5 of 23 | Official leaderboard | |
| FrontierMath | 31% | #18 of 70 | Epoch AI | |
| DTBench | 83.5% | #46 of 178 | Official leaderboard | |
| GPQA Diamond | 87.6% | #59 of 219 | Epoch AI | |
| OTIS Mock AIME 2024-2025 | 88.6% | #66 of 202 | Epoch AI | |
| FrontierMath Tier 4 | 12.5% | #18 of 55 | Epoch AI | |
| BRUMO 2025 | 93.3% | #16 of 42 | Official leaderboard | |
| HMMT Nov 2025 | 91.7% | #9 of 22 | Official leaderboard | |
| CritPt | 4.86% | #70 of 160 | Epoch AI | |
| ForecastBench | 62.1 | #39 of 86 | Official leaderboard | |
| SimpleBench | 53.2% | #48 of 100 | Official leaderboard | |
| ARC-AGI | 72.8% | #51 of 92 | Official leaderboard | |
| ARC-AGI-2 | 17.6% | #53 of 95 | Official leaderboard | |
| MathArena (Apex, IMO, USAMO) | 1.04% | #40 of 47 | Official leaderboard | |
| IMO-ProofBench | 7.10% | #10 of 10 | Official leaderboard | |
| Putnam 2025 | 58.0 | #4 of 4 | Official leaderboard |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Vibe Code Bench v1 | 24.6% | #4 of 17 | Official leaderboard | |
| WeirdML | 60.8% | #31 of 112 | Official leaderboard | |
| ALE-Bench | 1,192 | #41 of 121 | Epoch AI | |
| IOI (Vals) | 21.5% | #21 of 55 | Official leaderboard | |
| OTelBench | 14.5% | #7 of 14 | Official leaderboard | |
| GSO-Bench | 13.7% | #16 of 30 | Official leaderboard | |
| WebDev Arena | 1,392 | #69 of 105 | Epoch AI | |
| Vibe Code Bench | 24.6% | #77 of 115 | Official leaderboard | |
| Code Arena | 1,396 | #85 of 124 | Official leaderboard | |
| SWE-bench Verified | 68% | #25 of 32 | Epoch AI | |
| Image-to-WebDev Arena | 1,421 | #49 of 55 | Official leaderboard | |
| FrontierCS | 20.6 | #13 of 14 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Step Race | 5 | #3 of 59 | Official leaderboard | |
| SnakeBench | 33 | #20 of 183 | Official leaderboard | |
| ITBench | 55.2% | #2 of 13 | Official leaderboard | |
| GAIA | 75.8% | #11 of 57 | Official leaderboard | |
| LLM Chess | 948 | #26 of 124 | Official leaderboard | |
| Pencil Puzzle Bench | 6.70% | #19 of 53 | Official leaderboard | |
| Chess Puzzles | 42% | #32 of 83 | Epoch AI | |
| Terminal-Bench 2.0 | 47.6% | #18 of 43 | Official leaderboard | |
| tau2-bench | 80.2% | #33 of 69 | Model card | |
| Toolathlon | 37% | #26 of 50 | Official leaderboard | |
| Agent Poker Bench | 1,039 | #9 of 17 | Official leaderboard | |
| Mystery Game Puzzles | 19% | #47 of 80 | Epoch AI | |
| DeepResearch Bench | 42.8% | #19 of 24 | Epoch AI | |
| LMArena Search Arena | 1,184 | #28 of 35 | Official leaderboard | |
| MCP Atlas | 50.1% | #32 of 34 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CaseLaw v2 | 73.4% | #2 of 53 | Official leaderboard | |
| Vals Finance Agent v1 | 56.5% | #2 of 47 | Official leaderboard | |
| MedCode | 52.7% | #13 of 96 | Official leaderboard | |
| MedScribe | 88.1% | #14 of 98 | Official leaderboard | |
| LegalBench | 85.7% | #20 of 138 | Official leaderboard | |
| TaxEval v2 | 74.9% | #26 of 129 | Official leaderboard | |
| ProfBench | 54.9% | #19 of 72 | Official leaderboard | |
| Vals Finance Agent | 55.3% | #13 of 49 | Official leaderboard | |
| PRBench | 49.3% | #13 of 44 | Official leaderboard | |
| CorpFin v2 | 63.8% | #41 of 119 | Official leaderboard | |
| Vending-Bench | $2,380 | #5 of 9 | Official leaderboard | |
| FailureSensorIQ | 62.9% | #15 of 24 | Official leaderboard | |
| MortgageTax | 61.4% | #56 of 86 | Official leaderboard | |
| Vending-Bench 2 | $1,473 | #50 of 66 | Official leaderboard | |
| Vending-Bench Arena | $477 | #30 of 32 | Official leaderboard |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MMMU | 85.4% | #1 of 136 | Official leaderboard | |
| MMMU-Pro | 79% | #6 of 57 | Official leaderboard | |
| VPCT | 58.7% | #5 of 28 | Epoch AI | |
| EnigmaEval | 11.2% | #12 of 41 | Epoch AI | |
| RoboRewardBench | 0.90 | #6 of 19 | Official leaderboard | |
| Vision Arena | 1,251 | #44 of 136 | Official leaderboard | |
| VISTA | 43.8% | #28 of 54 | Official leaderboard | |
| ZeroBench | 6.00% | #40 of 66 | Official leaderboard | |
| Math Kangaroo 2025 (MathArena) | 76.9% | #15 of 22 | Official leaderboard | |
| OCR Arena | 1,367 | #9 of 11 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| AIR-Bench 2024 | 86.2% | #9 of 82 | Official leaderboard | |
| HELM Safety | 96.8 | #14 of 75 | Official leaderboard | |
| Anthropic Red Team | 99.5% | #19 of 75 | Official leaderboard | |
| StrongREJECT | 96.7% | #6 of 22 | Model card | |
| SimpleSafetyTests | 99.8% | #26 of 75 | Official leaderboard | |
| FORTRESS | 25.7% | #34 of 66 | Official leaderboard | |
| PCT (Polarized Contrastive Pairs) | 47.9% | #29 of 46 | Official leaderboard |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LiveBench | 78.8% | #2 of 48 | Epoch AI | |
| ACCoRD | 71.2 | #44 of 150 | Official leaderboard | |
| Conceptual Reasoning Index | 57.3 | #45 of 149 | Official leaderboard | |
| LMCA | 43.9% | #44 of 139 | Epoch AI | |
| HELM Capabilities | 65.6 | #31 of 60 | Official leaderboard | |
| CAIS Risk Index | 51.8 | #34 of 65 | Official leaderboard | |
| CAIS Text Capabilities Index | 30.7 | #35 of 66 | Official leaderboard | |
| CAIS Vision Capabilities Index | 47.4 | #33 of 53 | Official leaderboard | |
| SAGE (Vals) | 43.2% | #51 of 81 | Official leaderboard |
OpenAI's models over time
OpenAI's models in color, against every other model on the Epoch Capabilities Index, by release date.
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.
Its system card
Every result in the longer documents the lab published with it. 3 of them aren't in the launch post.
Robustness
- StrongREJECT 96.7