gpt-oss-120b ranks first on 5 of the 180 benchmarks with five or more models scored.
Open weights Mid-size
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 | |
|---|---|---|---|---|
| MMLU | 90% | #15 of 131 | Lab model card | |
| BOUQuET | 31.3 | #3 of 18 | Official leaderboard | |
| SnorkelSpatial | 52.7% | #5 of 27 | Official leaderboard | |
| Vectara Hallucination Leaderboard (HHEM) | 2.40% | #33 of 148 | Official leaderboard | |
| MMLU-Pro | 80.8% | #54 of 212 | Official leaderboard | |
| SuperGPQA | 51.9% | #20 of 59 | Official leaderboard | |
| LLM Confabulation Leaderboard | 15.7% | #25 of 58 | Official leaderboard | |
| IFEval | 83.6% | #28 of 60 | Official leaderboard | |
| Phare | 66.1% | #32 of 62 | Official leaderboard | |
| Wer wird Millionär? LLM Benchmark | 3111.0 | #40 of 77 | Official leaderboard | |
| LMArena (Elo) | 1,352 | #176 of 335 | Official leaderboard | |
| Global PIQA | 69.5% | #24 of 44 | Official leaderboard | |
| SEA-HELM | 66.1 | #40 of 71 | Official leaderboard | |
| MMMLU | 81.3% | #41 of 71 | Lab model card | |
| AraGen | 43.2% | #25 of 42 | Official leaderboard | |
| Last Translation Benchmark | 2.31 | #25 of 36 | Official leaderboard | |
| Humanity's Last Exam | 19% | #107 of 154 | Lab model card | |
| SimpleQA | 16.8% | #42 of 60 | Lab model card | |
| EQ-Bench 3 | 1,073 | #58 of 78 | Official leaderboard | |
| MultiChallenge | 45.3% | #32 of 42 | Official leaderboard | |
| StudyArena Text | 1,196 | #65 of 85 | Official leaderboard | |
| PersonQA | 29.8% | #7 of 9 | Lab model card | |
| SimpleQA Verified | 13.1% | #44 of 56 | Official leaderboard | |
| SimpleQA Hallucination Rate | 78.2% | #10 of 12 | Lab model card | |
| HLE Extended | 8.10% | #26 of 31 | Official leaderboard | |
| Fiction.LiveBench | 44.4% | #48 of 57 | Fiction.live leaderboard, via Epoch AI | |
| FACTS Parametric | 10.7% | #38 of 45 | Official leaderboard | |
| IndicGenBench | 39.1 | #11 of 13 | Official leaderboard | |
| BullshitBench V2 | 12% | #110 of 129 | Official leaderboard | |
| Vectara Hallucination Leaderboard (2025 dataset) | 14.2% | #90 of 104 | Official leaderboard | |
| Longform Creative Writing (EQ-Bench) | 35.8 | #104 of 119 | Official leaderboard | |
| Judgemark v2.1 | 54.8 | #37 of 42 | Official leaderboard | |
| MultiNRC | 15.2% | #34 of 38 | Official leaderboard | |
| SkateBench v2 | 16.4% | #19 of 21 | Official leaderboard | |
| Creative Writing v3 | 47.9 | #111 of 120 | Official leaderboard | |
| FACTS Search | 44.5% | #29 of 31 | Official leaderboard | |
| Lech Mazur Creative Writing | -3.17 | #47 of 50 | Official leaderboard | |
| AA-LCR Extended | 25.8% | #47 of 49 | Official leaderboard | |
| BullshitBench | 3.64% | #106 of 107 | Official leaderboard | |
| Judgemark v4 | 15.5 | #40 of 40 | Official leaderboard | |
| FACTS Grounding v2 | 20.1% | #25 of 25 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| AIME 2024 | 96.6% | #1 of 88 | Lab model card | |
| AIME 2025 | 97.9% | #15 of 143 | Lab model card | |
| MATH-500 | 94.8% | #6 of 38 | Official leaderboard | |
| IneqMath | 23.5% | #6 of 34 | Official leaderboard | |
| GPQA Main | 68.4% | #12 of 60 | Official leaderboard | |
| HiPhO | 62.2% | #9 of 36 | Official leaderboard | |
| HMMT 2025 | 90% | #14 of 53 | Official leaderboard | |
| SciArena | 1,037 | #10 of 34 | Official leaderboard | |
| OTIS Mock AIME 2024-2025 | 88.9% | #71 of 212 | Epoch AI run | |
| SMT 2025 | 87.7% | #15 of 40 | Official leaderboard | |
| ChemBench | 63.1% | #9 of 24 | Official leaderboard | |
| CMIMC 2025 | 85% | #13 of 33 | Official leaderboard | |
| BRUMO 2025 | 92.5% | #19 of 42 | Official leaderboard | |
| GPQA Diamond | 75.8% | #110 of 229 | Epoch AI run | |
| HMMT Nov 2025 | 90% | #11 of 22 | Official leaderboard | |
| DTBench | 76.3% | #103 of 180 | Official leaderboard | |
| A^3-Bench | 47.2% | #5 of 7 | Official leaderboard | |
| MathArena (Apex, IMO, USAMO) | 1.04% | #38 of 47 | Official leaderboard | |
| MathArena Apex Shortlist | 45.2% | #32 of 38 | Official leaderboard | |
| CORE-Bench | 11.1% | #15 of 17 | Official leaderboard | |
| SimpleBench | 22.1% | #97 of 106 | Official leaderboard | |
| Surface Evolver Bench | 25% | #23 of 25 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Multi-LCB | 67.8% | #1 of 11 | Official leaderboard | |
| AIRS-Bench | 0.40 | #1 of 5 | Official leaderboard | |
| Codeforces rating | 2,622 | #10 of 47 | Lab model card | |
| Spider 2.0 | 65.5% | #5 of 23 | Official leaderboard | |
| ArtifactsBench | 56.9% | #5 of 20 | Official leaderboard | |
| WeirdML | 48.2% | #50 of 112 | Official leaderboard | |
| ALE-Bench | 1,119 | #91 of 126 | Official leaderboard | |
| OpenAI PRs | 37% | #9 of 12 | Lab model card | |
| AlgoTune | 1.41× | #14 of 18 | Official leaderboard | |
| SWE-Bench Pro | 16.2% | #19 of 24 | Official leaderboard | |
| Aider Polyglot | 44.4% | #24 of 30 | Lab model card | |
| SWE-bench Verified | 26% | #50 of 60 | Official leaderboard | |
| DesignArena Game Dev | 982 | #126 of 143 | Official leaderboard | |
| SWE Atlas Codebase QnA Extended | 0% | #34 of 38 | Official leaderboard | |
| DesignArena Website | 961 | #138 of 151 | Official leaderboard | |
| PaperBench | 11% | #14 of 15 | Lab model card | |
| DesignArena UI Component | 918 | #131 of 139 | Official leaderboard | |
| DesignArena Code | 953 | #136 of 143 | Official leaderboard | |
| DesignArena 3D Design | 884 | #131 of 136 | Official leaderboard | |
| DesignArena Data Visualization | 985 | #114 of 118 | Official leaderboard | |
| APEX-SWE | 3.30% | #65 of 65 | Official leaderboard | |
| SWE Atlas v1.1 | 8.80% | #28 of 28 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| OdysseyArena AI Trading | 23.3% | #1 of 8 | Official leaderboard | |
| MCP-Bench | 0.69 | #3 of 15 | Official leaderboard | |
| TextQuests | 13.8 | #6 of 28 | Official leaderboard | |
| OdysseyArena Turn on Lights | 27.5% | #2 of 8 | Official leaderboard | |
| tau-bench | 67.8% | #7 of 23 | Lab model card | |
| SnakeBench | 26 | #69 of 188 | Official leaderboard | |
| LisanBench | 1143.7 | #53 of 133 | Official leaderboard | |
| BrowseComp-Plus | 42.9% | #3 of 7 | Official leaderboard | |
| VAKRA | 34.2% | #2 of 4 | Official leaderboard | |
| tau-bench Retail | 67.8% | #1 of 2 | Lab model card | |
| MCP-Universe | 25.5% | #16 of 30 | Official leaderboard | |
| Elimination Game | 4 | #32 of 56 | Official leaderboard | |
| Step Race | 2 | #35 of 59 | Official leaderboard | |
| lmgame-Bench | 22.2 | #14 of 23 | Official leaderboard | |
| BalatroBench | 6.27 | #10 of 16 | Official leaderboard | |
| SnorkelWordle | 81.6% | #12 of 19 | Official leaderboard | |
| Meta Agents Research Environments (ARE) / Gaia2 | 13.7% | #9 of 13 | Official leaderboard | |
| ScienceBoard | 7.04% | #11 of 15 | Official leaderboard | |
| OdysseyArena Energy Dispatch | 0% | #6 of 8 | Official leaderboard | |
| OdysseyArena Repo System | 18.3% | #6 of 8 | Official leaderboard | |
| Agent Poker Bench | 1,015 | #13 of 17 | Official leaderboard | |
| PACT | 1,496 | #20 of 26 | Official leaderboard | |
| LLM Chess | -217 | #97 of 123 | Official leaderboard | |
| PinchBench | 59.7% | #40 of 48 | Official leaderboard | |
| Terminal-Bench | 18.7% | #15 of 18 | tbench.ai, via Epoch AI | |
| PinchBench 2.0 | 44.8% | #50 of 57 | Official leaderboard | |
| MCPMark | 4.72% | #26 of 28 | Official leaderboard | |
| Terminal-Bench 2.0 | 18.7% | #40 of 43 | Official leaderboard | |
| CivBench | 1,422 | #18 of 19 | Official leaderboard | |
| CompileBench | 60% | #19 of 20 | Official leaderboard | |
| RuneBench | 1.45 | #74 of 76 | Official leaderboard | |
| Mystery Game Puzzles | 2.00% | #84 of 84 | Epoch AI run | |
| AutomationBench | 0.15% | #42 of 42 | Official leaderboard | |
| Collaborative Intelligence Benchmark | 19.1% | #18 of 18 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LEXam-hard | 37.6% | #4 of 15 | Official leaderboard | |
| LEXam | 51.7% | #11 of 30 | Official leaderboard | |
| ProfBench | 50% | #29 of 72 | Official leaderboard | |
| LMBuild | 51.3% | #8 of 18 | Official leaderboard | |
| Vals Finance Agent v1 | 28.3% | #26 of 47 | Official leaderboard | |
| TaxEval v2 | 71.6% | #73 of 129 | Official leaderboard | |
| FinQA (Snorkel OpenEnv environment) | 66.6% | #13 of 22 | Official leaderboard | |
| SnorkelFinance | 66.6% | #13 of 22 | Official leaderboard | |
| CorpFin v2 | 58.2% | #79 of 119 | Official leaderboard | |
| HealthBench | 57.6% | #21 of 31 | Lab model card | |
| LegalBench | 75.9% | #104 of 138 | Official leaderboard | |
| PRBench | 40.2% | #37 of 48 | Official leaderboard | |
| BRIDGE | 39.0 | #32 of 39 | Official leaderboard | |
| FailureSensorIQ | 50.5% | #20 of 24 | Official leaderboard | |
| Physical-Engineering's Last Exam | 0.05 | #15 of 18 | Official leaderboard | |
| Finance Reasoning (Snorkel) | 31.6% | #18 of 21 | Official leaderboard | |
| HealthBench Hard | 30% | #18 of 21 | Lab model card | |
| APEX-Agents | 4.70% | #7 of 8 | Official leaderboard | |
| SnorkelUnderwrite | 30% | #24 of 27 | Official leaderboard | |
| Vals Finance Agent | 21.5% | #44 of 49 | Official leaderboard | |
| FoodTruck Bench | $92 | #54 of 59 | Official leaderboard | |
| StockBench | 13.0 | #12 of 13 | Official leaderboard | |
| HealthBench Consensus | 89.9% | #16 of 17 | Lab model card | |
| CaseLaw v2 | 48.8% | #50 of 53 | Official leaderboard | |
| Vending-Bench 2 | $-22 | #65 of 67 | Official leaderboard | |
| APEX-1 | 46.4% | #70 of 71 | Official leaderboard | |
| APEX-Agents 1.1 | 8.30% | #65 of 65 | Official leaderboard | |
| TRLawBench | 4.78 | #34 of 34 | Official leaderboard | |
| PatientAgentBench | 3.45 | #8 of 8 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CaP-X (CaP-Bench) | 19.8% | #4 of 10 | Official leaderboard | |
| MineBench | 777 | #72 of 75 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| HELM Safety | 98.1 | #3 of 75 | Official leaderboard | |
| FORTRESS | 8.24% | #3 of 73 | Official leaderboard | |
| MolecularIQ | 47.5% | #1 of 17 | Official leaderboard | |
| AIR-Bench 2024 | 88% | #5 of 82 | Official leaderboard | |
| Spiral-Bench v1.0 | 81.4 | #3 of 13 | Official leaderboard | |
| Spiral-Bench v1.2 | 57.5 | #5 of 21 | Official leaderboard | |
| SimpleSafetyTests | 100% | #18 of 75 | Official leaderboard | |
| LiveSecBench | 66.6 | #9 of 34 | Official leaderboard | |
| Anthropic Red Team | 99.5% | #21 of 75 | Official leaderboard | |
| Tacit Knowledge and Troubleshooting | 74% | #11 of 30 | Lab model card | |
| Capture the Flag (Collegiate) | 33% | #5 of 13 | Lab model card | |
| Multimodal Troubleshooting Virology | 55% | #11 of 27 | Lab model card | |
| Lakera B3 (Backbone Breaker Benchmark) | 55.6% | #16 of 35 | Official leaderboard | |
| DTap Prompt Injection | 28.1% | #4 of 8 | Official leaderboard | |
| BioTIER Refuse | 45.4% | #36 of 69 | Official leaderboard | |
| Agentic BAIM-LLM Evaluation (ABLE) | 51.7% | #10 of 18 | Official leaderboard | |
| BioTIER Permit | 99.2% | #41 of 69 | Official leaderboard | |
| Molecular Biology Capabilities Test | 46.4% | #55 of 90 | Official leaderboard | |
| World-Class Bio | 40.7% | #46 of 74 | Official leaderboard | |
| DTap Direct Misuse | 45.2% | #5 of 8 | Official leaderboard | |
| Human Pathogen Capabilities Test | 45.8% | #57 of 86 | Official leaderboard | |
| Human Pathogen Capabilities Test v2 (HPCT v2) | 44.6% | #50 of 71 | Official leaderboard | |
| Capture the Flag (Professional) | 14% | #15 of 21 | Lab model card | |
| nocot-bench (NCRI) | 59.0 | #144 of 185 | Official leaderboard | |
| KORA | 22.6% | #35 of 42 | Official leaderboard | |
| ProtocolQA Open-Ended | 27% | #23 of 27 | Lab model card | |
| DistressBench | 63.7% | #36 of 42 | Official leaderboard | |
| TroubleshootingBench | 25% | #18 of 20 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| HELM Capabilities | 77.0 | #9 of 60 | Official leaderboard | |
| ACCoRD | 67.5 | #61 of 152 | Official leaderboard | |
| Conceptual Reasoning Index | 38.9 | #102 of 151 | Official leaderboard | |
| LMCA | 22.1% | #108 of 147 | Official leaderboard |
The Self-Reported Index fits one score to the 16 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 133.8 to 139.8. 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 | 62.4% SWE-Bench Verified, high reasoning (Table 3; n=477 Preparedness chart prints 62%) | 26% Leaderboard, mini-SWE-agent | +36 pts | OpenAI evaluates a fixed 477-task subset, while the board uses mini-SWE-agent on the 500-task Verified benchmark. |
| AIME 2025 | 97.9% AIME 2025 (tools) | 90% Leaderboard, high | +7.9 pts | |
| GPQA Diamond | 80.1% GPQA Diamond (without tools) | 75.8% Epoch AI run, high | +4.3 pts | |
| Aider Polyglot | 44.4% Aider Polyglot, high reasoning (Table 3) | 41.8% Leaderboard, high, aider, diff edits | +2.6 pts |
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.
Every result in the longer documents the lab published with it. 22 of them aren't in the launch post.