Introducing GPT-6 Sol and Luna 22 Sep 2026
GPT-6 Sol
GPT-6 Sol ranks first on 6 of the 168 benchmarks it has scores on.
177
Benchmarks with a score
6
Ranked first, out of five or more models
26
In the top three
OpenAI Sep 2026
GPT-6 Sol
Closed weights Mid-size
163.8
#9 of 271
Epoch Capabilities Index
- Terminal-Bench 4.0 49.4% #7 /23
- ZeroBench 41% #3 /66
- HealthBench Professional 60.8% #13 /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 | |
|---|---|---|---|---|
| Longform Creative Writing (EQ-Bench) | 82.8 | #8 of 119 | Official leaderboard | |
| Creative Writing v3 | 80.3 | #10 of 120 | Official leaderboard | |
| FACTS Benchmark Suite | 68.4% | #4 of 40 | Official leaderboard | |
| LMArena (Elo) | 1,458 | #50 of 333 | Official leaderboard | |
| FACTS Parametric | 75.6% | #8 of 45 | Official leaderboard | |
| SimpleQA Verified | 64.4% | #10 of 56 | Official leaderboard | |
| Extended NYT Connections | 90.1% | #21 of 97 | Official leaderboard | |
| FACTS Search V2 | 86.2% | #5 of 21 | Official leaderboard | |
| FACTS Parametric V2 | 76.1% | #5 of 20 | Official leaderboard | |
| FACTS Multimodal | 43.1% | #10 of 39 | Official leaderboard | |
| BullshitBench V2 | 59% | #33 of 122 | Official leaderboard | |
| Vectara Hallucination Leaderboard (2025 dataset) | 6.50% | #29 of 104 | Official leaderboard | |
| Global MMLU Lite | 94.1% | #12 of 43 | Official leaderboard | |
| BullshitBench | 41.8% | #35 of 107 | Official leaderboard | |
| Factuality Heavy prompts | 5.40% | #1 of 3 | Model card | |
| High Stakes factuality prompts | 4.10% | #1 of 3 | Model card | |
| HLE-Diamond | 43.2% | #7 of 18 | Official leaderboard | |
| OpenAI user-flagged hallucination eval | 13.5% | #3 of 4 | Model card |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| DTBench | 95.5% | #6 of 178 | Official leaderboard | |
| GPQA Diamond | 94.3% | #9 of 219 | Epoch AI | |
| OTIS Mock AIME 2024-2025 | 100% | #9 of 202 | Epoch AI | |
| CritPt | 30.9% | #8 of 160 | Epoch AI | |
| FrontierMath Tiers 1-3 v2 | 89.8% | #6 of 89 | Epoch AI | |
| ARC-AGI-2 | 89.6% | #8 of 95 | Official leaderboard | |
| FrontierMath Tier 4 v2 | 90% | #6 of 71 | Epoch AI | |
| ProofBench | 83% | #14 of 83 | Epoch AI | |
| SimpleBench | 73.1% | #17 of 100 | Official leaderboard | |
| ARC-AGI | 95.5% | #16 of 92 | Official leaderboard | |
| FrontierPhysics | 11.3% | #5 of 19 | Official leaderboard | |
| Leipzig Benchmark | 86% | #5 of 19 | Official leaderboard | |
| ProofBench v1.1 | 83% | #15 of 51 | Official leaderboard | |
| FrontierMath Erdős | 1.47% | #5 of 14 | Epoch AI | |
| LatchBio Capabilities v1.0 | 46.7% | #10 of 27 | Official leaderboard | |
| BixBench3 | 47.0 | #8 of 20 | Official leaderboard | |
| MysteryMechanism | 30.2% | #14 of 29 | Official leaderboard |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ALE-Bench | 2,462 | #4 of 121 | Epoch AI | |
| WebDev Arena | 1,689 | #6 of 105 | Epoch AI | |
| Code Arena | 1,678 | #11 of 124 | Official leaderboard | |
| Vibe Code Bench | 87.8% | #11 of 115 | Official leaderboard | |
| Code Migration | 57.2% | #9 of 78 | Official leaderboard | |
| Image-to-WebDev Arena | 1,660 | #7 of 55 | Official leaderboard | |
| FrontierCode | 49.3% | #9 of 46 | Epoch AI | |
| Vals RSI Index | 0.28 | #5 of 25 | Official leaderboard | |
| OpenDesign Arena | 89.9 | #4 of 19 | Official leaderboard | |
| Next.js Agent Evals | 97% | #3 of 14 | Official leaderboard | |
| IOI v2 (Vals) | 82.6% | #10 of 45 | Official leaderboard | |
| FrontierCode 2.0 | 34.3% | #6 of 26 | Official leaderboard | |
| WeirdML v3 | 19.7% | #8 of 25 | Official leaderboard | |
| DeepSWE | 68.8% | #20 of 51 | Model card | |
| Brokk Power Ranking | 8.20 | #9 of 19 | Official leaderboard | |
| FrontierCode 1.1 | 49.3% | #8 of 12 | Model card | |
| Internal Research Debugging Evaluation | 64.2% | #4 of 6 | Model card | |
| KernelGen 1P | 38.1% | #3 of 4 | Model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LLM Chess | 1,545 | #7 of 124 | Official leaderboard | |
| LisanBench | 8222.0 | #17 of 133 | Official leaderboard | |
| Mystery Game Puzzles | 56% | #11 of 80 | Epoch AI | |
| Game Arena Word Art | 1,405 | #5 of 35 | Official leaderboard | |
| Chess Puzzles | 56% | #13 of 83 | Epoch AI | |
| Game Arena Coin Game | 438 | #6 of 35 | Official leaderboard | |
| LMArena Search Arena | 1,228 | #7 of 35 | Official leaderboard | |
| AndroidWorld | 100% | #2 of 10 | Official leaderboard | |
| Agent Arena | 0.08 | #11 of 53 | Official leaderboard | |
| Game Arena Word Association | 334 | #8 of 36 | Official leaderboard | |
| EBR-bench | 53.3% | #7 of 28 | Epoch AI | |
| ARC-AGI-3 | 23% | #7 of 24 | Official leaderboard | |
| Hermes Bench | 73.5% | #5 of 17 | Official leaderboard | |
| Terminal-Bench 4.0 | 49.4% | #7 of 23 | Official leaderboard | |
| Game Arena Ultimate Tic-Tac-Toe | 392 | #11 of 36 | Official leaderboard | |
| Game Arena Chess | 1,007 | #12 of 39 | Official leaderboard | |
| Kaggle Game Arena | 331 | #14 of 39 | Official leaderboard | |
| OSWorld 2.0 | 64.4% | #7 of 19 | Model card | |
| Game Arena Bargaining | 210 | #14 of 36 | Official leaderboard | |
| Game Arena Lines of Action | 551 | #14 of 36 | Official leaderboard | |
| SpireClimb | 7.50% | #3 of 7 | Epoch AI | |
| SpireClimb StS1 A20 Heart | 1.67% | #3 of 7 | Epoch AI | |
| SpireClimb StS2 A10 | 13.3% | #3 of 7 | Epoch AI | |
| Game Arena Clobber | 314 | #16 of 36 | Official leaderboard | |
| Game Arena Checkers | 932 | #19 of 38 | Official leaderboard | |
| Game Arena Reversi | 504 | #18 of 36 | Official leaderboard | |
| Game Arena Four in a Row | 341 | #20 of 38 | Official leaderboard | |
| Game Arena Go | 905 | #19 of 36 | Official leaderboard | |
| AutomationBench | 33.2% | #22 of 41 | Model card | |
| Game Arena Dark Hex | 246 | #21 of 38 | Official leaderboard | |
| Game Arena Nine Men's Morris | 568 | #21 of 36 | Official leaderboard | |
| Game Arena Dots and Boxes | 817 | #23 of 37 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| PolicyBench | 95.2% | #1 of 50 | Official leaderboard | |
| Vending-Bench 2 | $14,428 | #3 of 66 | Official leaderboard | |
| TaxCalcBench v2 | 72% | #2 of 26 | Official leaderboard | |
| Agents' Last Exam | 56.4% | #2 of 25 | Model card | |
| Vending-Bench Arena | $10,530 | #5 of 32 | Official leaderboard | |
| Excel Modeling Benchmark | 71.5% | #12 of 76 | Official leaderboard | |
| HealthBench Consensus | 96.2% | #3 of 17 | Model card | |
| Clipping Bench | $1,240 | #4 of 19 | Official leaderboard | |
| Quant-Arena | 51.1 | #2 of 9 | Official leaderboard | |
| Thrive Holdings helpdesk ticket resolution benchmark | 55.8% | #4 of 15 | Official leaderboard | |
| MedCode | 47.1% | #35 of 96 | Official leaderboard | |
| Holdings-IndividualTaxBench | 72.9% | #5 of 13 | Official leaderboard | |
| CAD Arena | 0.53 | #5 of 12 | Official leaderboard | |
| APEX-Agents | 54.3% | #25 of 56 | Epoch AI | |
| EEBench | 56.3% | #12 of 26 | Official leaderboard | |
| MedScribe | 82% | #48 of 98 | Official leaderboard | |
| HealthBench Professional | 60.8% | #13 of 25 | Model card | |
| Short Form Bench | $2,394 | #5 of 9 | Official leaderboard | |
| Vals Finance Agent v2 | 49% | #48 of 80 | Official leaderboard | |
| Legal Research Bench | 28.9% | #54 of 88 | Official leaderboard | |
| Harvey LAB (Legal Agent Benchmark) | 1.67% | #54 of 80 | Official leaderboard | |
| Tax Agent Bench | 15% | #59 of 79 | Official leaderboard | |
| PRBench | 38.6% | #35 of 44 | Official leaderboard | |
| Public Benefits Bench | 56.6% | #40 of 50 | Official leaderboard | |
| HealthBench Hard | 30.1% | #17 of 21 | Model card | |
| HealthBench | 53.2% | #27 of 30 | Model card | |
| RCM-Bench-1 | 81.5% | #18 of 20 | Official leaderboard | |
| RealmPathology | 69.3% | #12 of 13 | Official leaderboard | |
| LegalReasonBench | 31.9% | #8 of 8 | Official leaderboard |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ZeroBench | 41% | #3 of 66 | Official leaderboard | |
| VoxelBench | 2,426 | #4 of 56 | Official leaderboard | |
| Vision Arena | 1,266 | #32 of 136 | Official leaderboard | |
| Furniture Assembly Benchmark | 58.3% | #9 of 35 | Epoch AI | |
| Code2Video Benchmark | 1,546 | #5 of 19 | Official leaderboard | |
| GDP.pdf | 26.4% | #11 of 39 | Epoch AI | |
| Blueprint-Bench 2 | 36.9% | #10 of 35 | Official leaderboard | |
| ParseBench | 68.9% | #17 of 42 | Official leaderboard | |
| Humanity's Sixth Sense | 31.2% | #14 of 29 | Official leaderboard | |
| RoboDojo-RC Tier 1 | 22% | #6 of 9 | Official leaderboard | |
| 1H-VideoQA | 66.3% | #16 of 20 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| DeepsecBench | 40.8 | #1 of 28 | Official leaderboard | |
| ProtocolQA Open-Ended | 44.4% | #1 of 25 | Model card | |
| CyberBench v1.1 (Vals) | 78% | #2 of 43 | Official leaderboard | |
| scBench-Long | 28.4% | #4 of 27 | Official leaderboard | |
| FORTRESS | 13.5% | #10 of 66 | Official leaderboard | |
| AAV Capsid Packaging Prediction (OpenAI) | 0.53 | #1 of 6 | Model card | |
| Coding Deception | 1.30% | #1 of 6 | Model card | |
| TxBench-Preclinical-Pharmacology | 63% | #5 of 28 | Official leaderboard | |
| SRE-Bench | 30.5% | #6 of 33 | Official leaderboard | |
| OpenAI Deployment Simulation | 0.08% | #1 of 5 | Model card | |
| Human Pathogen Capabilities Test v2 (HPCT v2) | 64.6% | #17 of 71 | Official leaderboard | |
| Tacit Knowledge and Troubleshooting | 79.2% | #7 of 29 | Model card | |
| TroubleshootingBench | 45.3% | #5 of 19 | Model card | |
| Prompt injection robustness | 99% | #2 of 7 | Model card | |
| ExploitBench | 82.6% | #6 of 19 | Model card | |
| TxBench-Oligo-Discovery | 43.1% | #9 of 28 | Official leaderboard | |
| SpatialBench-Long | 31.6% | #9 of 27 | Official leaderboard | |
| BioSecBench-Refusal | 37.9% | #8 of 22 | Official leaderboard | |
| OncologyBench | 20% | #11 of 30 | Official leaderboard | |
| Coronavirus-ACE2 Cell-Entry Screen | 0.46 | #2 of 5 | Model card | |
| Respecting Auto-Review | 0% | #2 of 5 | Model card | |
| Unintended Engagement with External Agent Messages | 26.3% | #2 of 5 | Model card | |
| BioSecBench-Surveillance | 47.7% | #11 of 27 | Official leaderboard | |
| TxBench-Antibody-Discovery | 43.3% | #12 of 28 | Official leaderboard | |
| Broken Search Tool | 4.92% | #3 of 7 | Model card | |
| Instruction hierarchy attacks | 100% | #3 of 7 | Model card | |
| BioSecBench-Function | 39.3% | #12 of 27 | Official leaderboard | |
| VariantBench | 38.7% | #12 of 27 | Official leaderboard | |
| MetagenomicsBench | 42% | #12 of 26 | Official leaderboard | |
| ExploitGym | 22.1% | #7 of 14 | Model card | |
| WellnessBench | 56% | #6 of 12 | Official leaderboard | |
| SEC-bench Pro | 68.8% | #5 of 10 | Model card | |
| CoT-Control | 23.2% | #4 of 8 | Model card | |
| Multimodal Troubleshooting Virology | 51.7% | #14 of 26 | Model card | |
| EpiBench | 22% | #16 of 27 | Official leaderboard | |
| Declining to Exploit a Honeypot During Difficult ExploitGym Problems | 0% | #3 of 5 | Model card | |
| Phage-plasmid Co-evolution | 12.9 | #3 of 5 | Model card | |
| Honeypot exploitation monitorability | 100% | #2 of 3 | Model card | |
| DistressBench | 68% | #31 of 42 | Official leaderboard | |
| ExploitBench - Internal Port | 5.50% | #3 of 4 | Model card | |
| Respecting Warnings | 28.1% | #4 of 5 | Model card | |
| SHP2 Protein Function Prediction | 0.27 | #4 of 5 | Model card | |
| CWE-bench | 52% | #11 of 12 | Official leaderboard | |
| MentalHealthBench | 48.8% | #4 of 4 | Model card | |
| Sandbox Bench | 4.50% | #4 of 4 | Model card | |
| Blocker Deception | 11.3% | #3 of 3 | Model card |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ACCoRD | 82.4 | #4 of 150 | Official leaderboard | |
| Conceptual Reasoning Index | 71.0 | #9 of 149 | Official leaderboard | |
| LMCA | 59.1% | #10 of 139 | Epoch AI | |
| Hermes Index | 44.1 | #5 of 17 | Official leaderboard | |
| Vals Science Index | 45% | #8 of 23 | Official leaderboard | |
| SAGE (Vals) | 44.8% | #43 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 cards
Every result in the longer documents the lab published with it. 42 of them aren't in the launch post.
Health
Hallucination
Safety
Robustness
Alignment
Monitorability
Bio and chem risk
Cyber risk
- ExploitBench 81.7
- ExploitBench - Internal Port 5.5
- SEC-bench Pro 66.3
- ExploitGym 22.1
- Sandbox Bench 4.5
AI R&D risk
Health
Hallucination
Safety
Robustness
Alignment
- Respecting Auto-Review 0
- Respecting Warnings 28.1
- Blocker Deception 11.3
- Broken Search Tool 5.2
Bio and chem risk
Cyber risk
- ExploitBench 82.6
- SEC-bench Pro 68.8