Introducing GPT-6.1 Sol 29 Sep 2026
GPT-6.1 Sol
GPT-6.1 Sol ranks first on 11 of the 193 benchmarks it has scores on.
198
Benchmarks with a score
11
Ranked first, out of five or more models
81
In the top three
OpenAI Sep 2026
GPT-6.1 Sol
Closed weights Mid-size
167.8
#4 of 271
Epoch Capabilities Index
- Terminal-Bench 4.0 58.2% #4 /23
- MultiChallenge 82% #1 /35
- HealthBench Professional 64.2% #8 /25
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 | |
|---|---|---|---|---|
| MultiChallenge | 82% | #1 of 35 | Official leaderboard | |
| Extended NYT Connections | 97.8% | #3 of 97 | Official leaderboard | |
| FACTS Parametric | 82.1% | #2 of 45 | Official leaderboard | |
| FACTS Benchmark Suite | 74.6% | #2 of 40 | Official leaderboard | |
| LMArena (Elo) | 1,481 | #24 of 333 | Official leaderboard | |
| FACTS Multimodal | 48.5% | #3 of 39 | Official leaderboard | |
| SimpleQA Verified | 73.8% | #5 of 56 | Official leaderboard | |
| FACTS Parametric V2 | 83.8% | #2 of 20 | Official leaderboard | |
| HLE-Diamond | 61.3% | #2 of 18 | Official leaderboard | |
| FACTS Grounding v2 | 78% | #2 of 15 | Official leaderboard | |
| FACTS Search V2 | 88.2% | #3 of 21 | Official leaderboard | |
| BullshitBench V2 | 65% | #27 of 122 | Official leaderboard | |
| EQ-Bench 4 | 1,272 | #8 of 32 | Official leaderboard | |
| Global MMLU Lite | 78.9% | #33 of 43 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| FrontierMath Tier 4 v2 | 100% | #1 of 71 | Epoch AI | |
| ARC-AGI-2 | 94.2% | #2 of 95 | Official leaderboard | |
| FrontierMath Tiers 1-3 v2 | 93.7% | #2 of 89 | Epoch AI | |
| GPQA Diamond | 95.4% | #5 of 219 | Epoch AI | |
| CritPt | 31.7% | #5 of 160 | Epoch AI | |
| SimpleBench | 82.9% | #5 of 100 | Official leaderboard | |
| ARC-AGI | 98.5% | #5 of 92 | Official leaderboard | |
| OTIS Mock AIME 2024-2025 | 100% | #11 of 202 | Epoch AI | |
| FrontierMath Erdős | 4.41% | #1 of 14 | Epoch AI | |
| ProofBench | 99% | #6 of 83 | Epoch AI | |
| DTBench | 94.2% | #13 of 178 | Official leaderboard | |
| ProofBench v1.1 | 99% | #6 of 51 | Official leaderboard | |
| FrontierPhysics | 16.1% | #3 of 19 | Official leaderboard | |
| ArXivMath | 94.2% | #1 of 6 | Official leaderboard | |
| BrokenArXiv | 89.7% | #1 of 6 | Official leaderboard | |
| MysteryMechanism | 46.4% | #5 of 29 | Official leaderboard | |
| CritPt v2 | 82% | #2 of 11 | Official leaderboard | |
| LatchBio Capabilities v1.0 | 50.6% | #6 of 27 | Official leaderboard | |
| Prophet Arena | 85.7 | #4 of 18 | Official leaderboard | |
| ArXivLean | 35.9% | #4 of 7 | Official leaderboard | |
| Terminal-Bench-Science | 57% | #6 of 9 | Model card |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ALE-Bench | 2,744 | #2 of 121 | Epoch AI | |
| WebDev Arena | 1,759 | #4 of 105 | Epoch AI | |
| Code Arena | 1,751 | #5 of 124 | Official leaderboard | |
| Code Migration | 65.1% | #5 of 78 | Official leaderboard | |
| IOI v2 (Vals) | 96.9% | #3 of 45 | Official leaderboard | |
| Vibe Code Bench | 88.9% | #9 of 115 | Official leaderboard | |
| WeirdML v3 | 38.3% | #2 of 25 | Official leaderboard | |
| Image-to-WebDev Arena | 1,702 | #5 of 55 | Official leaderboard | |
| ViBench 2.0 | 64.7% | #1 of 9 | Official leaderboard | |
| DeepSWE | 75.2% | #7 of 51 | Model card | |
| Code Migration v1.1 | 77.3% | #4 of 29 | Official leaderboard | |
| FrontierCode | 50.2% | #7 of 46 | Epoch AI | |
| OpenDesign Arena | 90.1 | #3 of 19 | Official leaderboard | |
| FrontierCode 2.0 | 39% | #5 of 26 | Official leaderboard | |
| GameDevBench | 70% | #2 of 10 | Official leaderboard | |
| Brokk Power Ranking | 19.9 | #4 of 19 | Official leaderboard | |
| RSI Bench | 37.0 | #2 of 8 | Official leaderboard | |
| KernelGen 1P | 60.4% | #1 of 4 | Model card | |
| Android Bench 2.0 | 29.3% | #3 of 11 | Official leaderboard | |
| Vals RSI Index | 0.27 | #7 of 25 | Official leaderboard | |
| MirrorCode | 61.1% | #8 of 28 | Epoch AI | |
| PostTrainBench 1.2 | 38.8% | #5 of 17 | Official leaderboard | |
| FrontierCS 2 | 35.4% | #7 of 16 | Official leaderboard | |
| SWE-Bench Pro V2 | 96.1% | #7 of 16 | Official leaderboard | |
| Web Clone Bench | 12.6% | #4 of 9 | Official leaderboard | |
| ScarfBench | 68.4% | #5 of 10 | Official leaderboard | |
| Internal Research Debugging Evaluation | 75.5% | #3 of 6 | Model card | |
| SWE-Marathon v1.1 | 42.5% | #13 of 17 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LLM Chess | 1,612 | #2 of 124 | Official leaderboard | |
| Mystery Game Puzzles | 80% | #2 of 80 | Epoch AI | |
| Chess Puzzles | 68% | #3 of 83 | Epoch AI | |
| LMArena Search Arena | 1,250 | #2 of 35 | Official leaderboard | |
| GAIA | 86.4% | #4 of 57 | Official leaderboard | |
| Game Arena Checkers | 1,212 | #3 of 38 | Official leaderboard | |
| Game Arena Dots and Boxes | 1,134 | #3 of 37 | Official leaderboard | |
| Game Arena Clobber | 655 | #3 of 36 | Official leaderboard | |
| Game Arena Lines of Action | 973 | #3 of 36 | Official leaderboard | |
| Game Arena Word Art | 1,968 | #3 of 35 | Official leaderboard | |
| Kaggle Game Arena | 436 | #4 of 39 | Official leaderboard | |
| Game Arena Word Association | 439 | #4 of 36 | Official leaderboard | |
| CivBench | 1,602 | #2 of 18 | Official leaderboard | |
| Agent Arena | 0.12 | #6 of 53 | Official leaderboard | |
| LisanBench | 8736.0 | #16 of 133 | Official leaderboard | |
| ARC-AGI-3 | 96.4% | #3 of 24 | Official leaderboard | |
| Game Arena Bargaining | 245 | #5 of 36 | Official leaderboard | |
| Game Arena Ultimate Tic-Tac-Toe | 497 | #5 of 36 | Official leaderboard | |
| SpireClimb StS1 A20 Heart | 20% | #1 of 7 | Epoch AI | |
| OSWorld 2.0 | 71.4% | #3 of 19 | Model card | |
| Terminal-Bench 4.0 | 58.2% | #4 of 23 | Official leaderboard | |
| Hermes Bench | 73.9% | #3 of 17 | Official leaderboard | |
| Game Arena Four in a Row | 573 | #7 of 38 | Official leaderboard | |
| Game Arena Go | 1,441 | #7 of 36 | Official leaderboard | |
| Game Arena Reversi | 784 | #7 of 36 | Official leaderboard | |
| EBR-bench | 54.3% | #6 of 28 | Epoch AI | |
| StarSkirmish Bench | 45.8% | #2 of 9 | Official leaderboard | |
| Game Arena Chess | 1,230 | #9 of 39 | Official leaderboard | |
| Game Arena Coin Game | 418 | #10 of 35 | Official leaderboard | |
| SpireClimb | 24.7% | #2 of 7 | Epoch AI | |
| SpireClimb StS2 A10 | 29.3% | #2 of 7 | Epoch AI | |
| Game Arena Nine Men's Morris | 766 | #11 of 36 | Official leaderboard | |
| Game Arena Dark Hex | 280 | #16 of 38 | Official leaderboard | |
| AutomationBench | 36.1% | #21 of 41 | Model card | |
| o11y-bench | 87.5% | #4 of 6 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Clipping Bench | $5,707 | #2 of 19 | Official leaderboard | |
| Short Form Bench | $2,949 | #1 of 9 | Official leaderboard | |
| TaxCalcBench v2 | 70% | #4 of 26 | Official leaderboard | |
| Excel Modeling Benchmark | 70.8% | #14 of 76 | Official leaderboard | |
| PolicyBench | 90.2% | #10 of 50 | Official leaderboard | |
| Holdings-IndividualTaxBench | 74.2% | #3 of 13 | Official leaderboard | |
| MedScribe | 86.5% | #23 of 98 | Official leaderboard | |
| HealthBench Consensus | 96% | #4 of 17 | Model card | |
| Vals Index v2.2 | 63.9% | #8 of 32 | Official leaderboard | |
| APEX-Agents | 60% | #15 of 56 | Epoch AI | |
| EEBench | 63.6% | #7 of 26 | Official leaderboard | |
| RWE-bench | 25% | #3 of 11 | Official leaderboard | |
| MedCode | 48.8% | #28 of 96 | Official leaderboard | |
| HealthBench Professional | 64.2% | #8 of 25 | Model card | |
| Excel Modeling Benchmark v1.1 | 71.1% | #14 of 37 | Official leaderboard | |
| Vals Finance Agent v2.1 | 61.8% | #14 of 35 | Official leaderboard | |
| PRBench | 47.9% | #18 of 44 | Official leaderboard | |
| Immigration Bench | 67.8% | #10 of 24 | Official leaderboard | |
| Legal Research Bench | 38.5% | #37 of 88 | Official leaderboard | |
| HealthBench Hard | 36.2% | #9 of 21 | Model card | |
| Harvey LAB (Legal Agent Benchmark) | 5.42% | #36 of 80 | Official leaderboard | |
| Vals Finance Agent v2 | 52% | #37 of 80 | Official leaderboard | |
| HealthBench | 58.5% | #14 of 30 | Model card | |
| Vals Legal Index | 68.7% | #19 of 33 | Official leaderboard | |
| Legora BAR | 102.0 | #10 of 17 | Official leaderboard | |
| Tax Agent Bench | 21.8% | #47 of 79 | Official leaderboard | |
| Vals Finance Index | 55.8% | #23 of 35 | Official leaderboard | |
| Public Benefits Bench | 59.3% | #35 of 50 | Official leaderboard | |
| RCM-Bench-1 | 87.4% | #14 of 20 | Official leaderboard | |
| Harvey LAB v1.1 (HLAB-Vals-1.1) | 6.25% | #25 of 33 | Official leaderboard | |
| MissionReport | 61% | #7 of 9 | Official leaderboard |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GDP.pdf | 35% | #1 of 39 | Epoch AI | |
| VoxelBench | 2,647 | #2 of 56 | Official leaderboard | |
| Vision Arena | 1,294 | #7 of 136 | Official leaderboard | |
| Furniture Assembly Benchmark | 80% | #3 of 35 | Epoch AI | |
| Humanity's Sixth Sense | 46.6% | #3 of 29 | Official leaderboard | |
| ClockBench | 74.4% | #4 of 37 | Official leaderboard | |
| Robo-bench | 28.3% | #1 of 7 | Official leaderboard | |
| ScienceVision | 0.49 | #2 of 11 | Official leaderboard | |
| DataCenterBench Tier 1 | 38.3% | #2 of 9 | Official leaderboard | |
| Dual-Arm Sim | 40.8 | #3 of 10 | Official leaderboard | |
| ParseBench | 69.7% | #13 of 42 | Official leaderboard | |
| RoboColiseum | 70.9% | #3 of 9 | Official leaderboard | |
| RoboDojo-RC Tier 1 | 41.7% | #3 of 9 | Official leaderboard | |
| OmniParseBench | 91.5 | #2 of 6 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SRE-Bench | 50.8% | #2 of 33 | Official leaderboard | |
| Tacit Knowledge and Troubleshooting | 88.5% | #2 of 29 | Model card | |
| DeepsecBench | 39.5 | #2 of 28 | Official leaderboard | |
| WellnessBench | 66% | #1 of 12 | Official leaderboard | |
| ExploitBench | 99.7% | #2 of 19 | Model card | |
| TxBench-Oligo-Discovery | 52.5% | #3 of 28 | Official leaderboard | |
| SpatialBench-Long | 39.9% | #3 of 27 | Official leaderboard | |
| ExploitGym | 35.1% | #2 of 18 | Official leaderboard | |
| BioSecBench-Function | 46% | #5 of 27 | Official leaderboard | |
| TroubleshootingBench | 48% | #4 of 19 | Model card | |
| PCT (Polarized Contrastive Pairs) | 37.3% | #10 of 46 | Official leaderboard | |
| BioSecBench-Surveillance | 50.8% | #6 of 27 | Official leaderboard | |
| VariantBench | 46% | #6 of 27 | Official leaderboard | |
| OncologyBench | 24% | #7 of 30 | Official leaderboard | |
| ProtocolQA Open-Ended | 40.7% | #6 of 25 | Model card | |
| MentalHealthBench | 57.9% | #1 of 4 | Model card | |
| EpiBench | 28.3% | #7 of 27 | Official leaderboard | |
| Broken Search Tool | 2.08% | #2 of 7 | Model card | |
| Instruction hierarchy attacks | 100% | #2 of 7 | Model card | |
| scBench-Long | 26.7% | #8 of 27 | Official leaderboard | |
| MetagenomicsBench | 45.7% | #8 of 26 | Official leaderboard | |
| Multimodal Troubleshooting Virology | 55.3% | #8 of 26 | Model card | |
| TxBench-Antibody-Discovery | 49.7% | #9 of 28 | Official leaderboard | |
| TxBench-Preclinical-Pharmacology | 59.7% | #9 of 28 | Official leaderboard | |
| DistressBench | 74.2% | #14 of 42 | Official leaderboard | |
| Coding Deception | 1.50% | #2 of 6 | Model card | |
| Misaligned Behavior in Realistic Work Environments | 1.94% | #1 of 3 | Model card | |
| CoT-Control | 44.8% | #3 of 8 | Model card | |
| SEC-bench Pro | 78.8% | #4 of 10 | Model card | |
| BioSecBench-Refusal | 37.6% | #9 of 22 | Official leaderboard | |
| Prompt injection robustness | 99% | #3 of 7 | Model card | |
| ExploitBench - Internal Port | 21.5% | #2 of 4 | Model card | |
| CheatBench | 49.3% | #6 of 10 | Official leaderboard | |
| Coronavirus-ACE2 Cell-Entry Screen | 0.42 | #3 of 5 | Model card | |
| OpenAI Deployment Simulation | 0.06% | #3 of 5 | Model card | |
| Respecting Auto-Review | 0% | #3 of 5 | Model card | |
| Respecting Warnings | 23.5% | #3 of 5 | Model card | |
| SHP2 Protein Function Prediction | 0.33 | #3 of 5 | Model card | |
| Declining to Exploit a Honeypot During Difficult ExploitGym Problems | 0% | #4 of 5 | Model card | |
| Phage-plasmid Co-evolution | 12.9 | #4 of 5 | Model card | |
| Unintended Engagement with External Agent Messages | 38% | #4 of 5 | Model card | |
| AAV Capsid Packaging Prediction (OpenAI) | 0.53 | #5 of 6 | Model card | |
| CyberBench v1.1 (Vals) | 39.3% | #41 of 43 | Official leaderboard | |
| Honeypot exploitation monitorability | 100% | #3 of 3 | Model card |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Conceptual Reasoning Index | 72.7 | #6 of 149 | Official leaderboard | |
| CAIS Text Capabilities Index | 61.4 | #3 of 66 | Official leaderboard | |
| LMCA | 62.8% | #7 of 139 | Epoch AI | |
| CAIS Vision Capabilities Index | 77.9 | #3 of 53 | Official leaderboard | |
| ACCoRD | 80.9 | #9 of 150 | Official leaderboard | |
| SAGE v1.1 (Vals) | 90.5% | #2 of 28 | Official leaderboard | |
| Vals Science Index | 59.6% | #2 of 23 | Official leaderboard | |
| Hermes Index | 57.1 | #2 of 17 | Official leaderboard | |
| LiveBench | 84.1% | #4 of 28 | Official leaderboard | |
| CAIS Risk Index | 39.9 | #13 of 65 | Official leaderboard | |
| SAGE (Vals) | 46.5% | #35 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. 43 of them aren't in the launch post.
Addendum: GPT-6.1 Sol System Card 29 Sep 2026
Health
Hallucination
Safety
Robustness
Alignment
- Respecting Auto-Review 0
- Respecting Warnings 23.5
- Declining to Exploit a Honeypot During Difficult ExploitGym Problems 0
- Coding Deception 1.5
- Broken Search Tool 2.08
- Misaligned Behavior in Realistic Work Environments 1.94
- Unintended Engagement with External Agent Messages 38
- OpenAI Deployment Simulation 0.056
Monitorability
Bio and chem risk
Cyber risk
- ExploitBench 99.7
- ExploitBench - Internal Port 21.5
- SEC-bench Pro 78.8
- ExploitGym 35.1