- Agentic terminal coding Frontier-Bench v0.1 43.3
- Knowledge work GDPval-AA v2 1861
- Novel problem-solving ARC-AGI-3 (high) 30.2
- Agentic search BrowseComp 90.8
- Multidisciplinary reasoning Humanity's Last Exam (with tools) 64.7
- Computer use OSWorld 2.0 70.6
- Agentic coding DeepSWE v1.1 68.8
- Agentic coding FrontierCode v1.1, Main 53.4
- Business workflows AutomationBench 26
- Legal Legal Agent Benchmark, Held-out 11.7
- Health HealthBench Professional 59.8
- Best of Biology BioMysteryBench (hard) / Biology BioMysteryBench (human solved) 90.1
- SWE-bench Pro 79.2
- SWE-bench Multilingual 89.5
- SWE-bench Multimodal 59.4
- AA-Briefcase 1720
- ARC-AGI-1 97.5
- ARC-AGI-2 90.4
Claude Opus 5
Claude Opus 5 ranks first on 55 of the 380 benchmarks with five or more models scored.
406
Benchmarks with a score
55
Ranked first, out of five or more models
121
In the top three
Anthropic Jul 2026
Claude Opus 5
Closed weights Flagship
164.7
#10 of 256
Epoch Capabilities Index
Self-Reported Index 166.8 #10 of 404
- DeepSWE 73.7% #3 /28
- AutomationBench 26.9% #12 /42
- Agents' Last Exam 32.2% #3 /31
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) | 86.3 | #1 of 119 | Official leaderboard | |
| Global MMLU Lite | 95.7% | #1 of 43 | Official leaderboard | |
| DesignArena ASCII Art | 1,353 | #2 of 77 | Official leaderboard | |
| EQ-Bench 4 | 1,365 | #1 of 32 | Official leaderboard | |
| INCLUDE | 89.8% | #1 of 31 | Lab model card | |
| LMArena (Elo) | 1,490 | #13 of 335 | Official leaderboard | |
| Humanity's Last Exam | 64.7% | #6 of 154 | Lab model card | |
| Creative Writing v3 | 81.3 | #6 of 120 | Official leaderboard | |
| Lech Mazur Creative Writing | 3.69 | #3 of 50 | Official leaderboard | |
| BullshitBench V2 | 80.2% | #10 of 129 | Official leaderboard | |
| Corp Speak Arena | 1,696 | #2 of 22 | Official leaderboard | |
| Continual Learning Exam | 47.7% | #1 of 11 | Official leaderboard | |
| HLE Extended | 61% | #3 of 31 | Official leaderboard | |
| Extended NYT Connections | 94.3% | #10 of 103 | Official leaderboard | |
| AA-LCR Extended | 73.9% | #7 of 49 | Official leaderboard | |
| Document Arena | 1,486 | #6 of 39 | Official leaderboard | |
| Judgemark v4 | 78.8 | #7 of 40 | Official leaderboard | |
| BullshitBench | 60% | #19 of 107 | Official leaderboard | |
| OpenAI MRCR v2 | 65.6% | #9 of 47 | Official leaderboard | |
| SimpleQA Verified | 63.3% | #11 of 56 | Official leaderboard | |
| Antidote | 1,055 | #10 of 36 | Official leaderboard | |
| FACTS Search V2 | 85.3% | #6 of 21 | Official leaderboard | |
| Global MMLU | 92.5% | #4 of 14 | Lab model card | |
| Hemingway-bench | 1,047 | #13 of 42 | Official leaderboard | |
| MILU | 92.1% | #4 of 11 | Lab model card | |
| StudyArena Text | 1,242 | #32 of 85 | Official leaderboard | |
| Joke Arena | 1,515 | #11 of 25 | Official leaderboard | |
| FACTS Parametric | 48.8% | #20 of 45 | Official leaderboard | |
| HLE-Diamond | 40.5% | #12 of 20 | Official leaderboard | |
| FACTS Parametric V2 | 48.8% | #17 of 27 | Official leaderboard | |
| ELI5 Arena | 1,436 | #13 of 18 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ProofBench | 78% | #1 of 59 | Official leaderboard | |
| DTBench | 97.9% | #4 of 180 | Official leaderboard | |
| ForecastBench | 0.11 | #3 of 82 | Official leaderboard | |
| NatureBench | 26.7% | #1 of 18 | Official leaderboard | |
| GPQA Diamond | 93.9% | #17 of 229 | Epoch AI run | |
| ARC-AGI-2 | 90.4% | #8 of 98 | Official leaderboard | |
| ProofBench v1.1 | 99% | #5 of 52 | Official leaderboard | |
| ExplorationBench | 89% | #1 of 10 | Official leaderboard | |
| SimpleBench | 80.6% | #12 of 106 | Official leaderboard | |
| ARC-AGI | 97.5% | #11 of 95 | Official leaderboard | |
| OTIS Mock AIME 2024-2025 | 98.9% | #25 of 212 | Epoch AI run | |
| BTF-3 | 0.12 | #1 of 8 | Official leaderboard | |
| BioMysteryBench | 79.3% | #4 of 30 | Official leaderboard | |
| PhysClosureBench | 6.29 | #1 of 7 | Official leaderboard | |
| FrontierMath Tiers 1-3 v2 | 85.6% | #15 of 86 | Official leaderboard | |
| LiveMathematicianBench | 42.7% | #3 of 17 | Official leaderboard | |
| ASI-Bench | 50.3 | #2 of 11 | Official leaderboard | |
| Human Behavior Bench | 80.2 | #2 of 11 | Official leaderboard | |
| FrontierMath Tier 4 v2 | 73.2% | #14 of 68 | Official leaderboard | |
| Science's Last Exam | 37.5 | #2 of 9 | Official leaderboard | |
| Terminal-Bench-Science | 30% | #4 of 17 | Official leaderboard | |
| Apodex Discovery | 2.74 | #2 of 8 | Official leaderboard | |
| MysteryMechanism | 37.4% | #8 of 31 | Official leaderboard | |
| LatchBio Capabilities v1.0 | 50% | #7 of 27 | Official leaderboard | |
| SpatialBench (LatchBio) | 72.5% | #2 of 7 | Lab model card | |
| FrontierPhysics | 9.50% | #7 of 22 | Official leaderboard | |
| ArXivMath | 90.8% | #3 of 9 | Lab model card | |
| FrontierChallenge | 17.5% | #4 of 11 | Official leaderboard | |
| Leipzig Benchmark | 75% | #8 of 19 | Official leaderboard | |
| scBench (LatchBio SingleCellBench) | 58.3% | #11 of 26 | Official leaderboard | |
| EurekaBench | 27.3 | #3 of 7 | Official leaderboard | |
| SciUniverse | 30.5% | #3 of 6 | Official leaderboard | |
| Into the Unknown | 0% | #5 of 9 | Official leaderboard | |
| PaperBenchX | 7.53% | #6 of 10 | Official leaderboard | |
| BixBench3 | 40.6 | #13 of 20 | Official leaderboard | |
| Riemann-bench | 60% | #3 of 4 | Lab model card | |
| FrontierMath Erdős | 0% | #12 of 15 | Official leaderboard | |
| Organic Chemistry V2 (Anthropic) | 61.6% | #2 of 2 | Lab model card | |
| IMO 2026 | 100% | #1 of 1 | Lab model card | |
| Protein Design (Anthropic) | 42.5% | #1 of 1 | Lab model card |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SWE-bench Verified | 96% | #1 of 155 | Lab model card | |
| IOI (Vals) | 91.7% | #1 of 55 | Official leaderboard | |
| AndroidBench | 91.8% | #1 of 32 | Official leaderboard | |
| WeirdML | 91.8% | #5 of 112 | Official leaderboard | |
| SWE Atlas | 63.2% | #1 of 22 | Official leaderboard | |
| APEX-SWE | 63.7% | #3 of 65 | Official leaderboard | |
| GBAEval | 79.6% | #1 of 21 | Official leaderboard | |
| SWE-bench Multilingual | 89.5% | #4 of 73 | Lab model card | |
| SWE-Bench Pro V2 | 99.4% | #1 of 18 | Official leaderboard | |
| SWE-Bench Pro V2 Hard | 98% | #1 of 18 | Official leaderboard | |
| DesignArena SVG | 1,307 | #6 of 103 | Official leaderboard | |
| WebDev Arena | 1,687 | #7 of 105 | arena.ai, via Epoch AI | |
| SWE-bench Verified Extended | 82% | #3 of 44 | Official leaderboard | |
| Code Arena | 1,691 | #9 of 128 | Official leaderboard | |
| SWE-Bench Pro | 79.2% | #6 of 81 | Lab model card | |
| DesignArena Data Visualization | 1,327 | #9 of 118 | Official leaderboard | |
| SWE Atlas Codebase QnA Extended | 40% | #3 of 38 | Official leaderboard | |
| DesignArena UI Component | 1,341 | #11 of 139 | Official leaderboard | |
| ALE-Bench | 2,622 | #10 of 126 | Official leaderboard | |
| SWE Atlas Test Writing | 62.2% | #2 of 22 | Official leaderboard | |
| PostTrainBench 1.1 | 45.6% | #1 of 11 | Official leaderboard | |
| Vibe Code Bench | 88.4% | #11 of 120 | Official leaderboard | |
| Code Migration | 57.5% | #8 of 78 | Official leaderboard | |
| MirrorCode | 87.8% | #4 of 39 | Epoch AI run | |
| Doc2Harness | 46.6% | #2 of 19 | Official leaderboard | |
| DeepSWE | 73.7% | #3 of 28 | Official leaderboard | |
| Image-to-WebDev Arena | 1,664 | #6 of 55 | Official leaderboard | |
| DesignArena Code | 1,318 | #16 of 143 | Official leaderboard | |
| DesignArena Game Dev | 1,331 | #16 of 143 | Official leaderboard | |
| Vibe Code Bench 1-100 | 28.5% | #3 of 26 | Official leaderboard | |
| DesignArena Website | 1,305 | #18 of 151 | Official leaderboard | |
| MiMo Code Bench | 68.6% | #1 of 8 | Official leaderboard | |
| Quadruped reinforcement learning task (Anthropic) | 31.3 | #1 of 8 | Lab model card | |
| SWE Refactor Bench | 47% | #1 of 8 | Official leaderboard | |
| DesignArena 3D Design | 1,324 | #18 of 136 | Official leaderboard | |
| DesignArena Full Stack | 1,285 | #7 of 49 | Official leaderboard | |
| DesignArena Agentic Game Dev | 1,204 | #4 of 27 | Official leaderboard | |
| Vals RSI Index | 0.33 | #4 of 27 | Official leaderboard | |
| DesignArena Android (Kotlin) | 1,244 | #7 of 46 | Official leaderboard | |
| DesignArena Godot Game Dev | 1,261 | #4 of 24 | Official leaderboard | |
| DesignArena Agentic Overall | 1,264 | #13 of 74 | Official leaderboard | |
| IOI v2 (Vals) | 84.3% | #8 of 45 | Official leaderboard | |
| ReactBench | 42% | #5 of 28 | Official leaderboard | |
| TasteVal | 128.6× | #4 of 20 | Official leaderboard | |
| GameCraft-Bench | 68.4% | #3 of 15 | Official leaderboard | |
| SWE-bench Science | 47.9% | #2 of 10 | Official leaderboard | |
| Vision2Web | 80.2% | #2 of 10 | Official leaderboard | |
| StepCodeBench | 63.9% | #1 of 5 | Official leaderboard | |
| τ^τ-bench | 23.9% | #1 of 5 | Official leaderboard | |
| Senior SWE-Bench | 34.7% | #5 of 23 | Official leaderboard | |
| Data Agent Benchmark | 95.4% | #4 of 18 | Official leaderboard | |
| SWE-Clar | 63% | #2 of 9 | Official leaderboard | |
| SWE-rebench | 63.4% | #3 of 13 | Official leaderboard | |
| PostTrainBench 1.2 | 41.7% | #5 of 21 | Official leaderboard | |
| ProgramBench | 4.50% | #6 of 25 | Official leaderboard | |
| ProgramBench almost solved | 37% | #6 of 25 | Official leaderboard | |
| ProgramBench tests passed | 74.7% | #6 of 25 | Official leaderboard | |
| DesignArena Frontend Web App | 1,253 | #13 of 53 | Official leaderboard | |
| SWE-bench CLI | 14% | #4 of 16 | Official leaderboard | |
| SWE-bench Multimodal | 59.4% | #3 of 12 | Lab model card | |
| Novel compiler task (Anthropic) | 80.9% | #1 of 4 | Lab model card | |
| MLS-Bench | 49.8% | #4 of 15 | Official leaderboard | |
| ScarfBench | 69.8% | #3 of 11 | Official leaderboard | |
| DAXBench v2 | 85.4% | #13 of 47 | Official leaderboard | |
| RSI-Exam | 46.4% | #5 of 18 | Official leaderboard | |
| FrontierCode 1.1 | 53.4% | #4 of 13 | Lab model card | |
| SWE-Together | 63.8% | #7 of 22 | Official leaderboard | |
| Legacy-Bench | 52% | #2 of 6 | Official leaderboard | |
| FrontierSWE | 52% | #9 of 25 | Official leaderboard | |
| WeirdML v3 | 19% | #10 of 27 | Official leaderboard | |
| FrontierCode 2.0 | 28.1% | #13 of 32 | Official leaderboard | |
| AutoResearchExam | 0.58 | #5 of 12 | Official leaderboard | |
| Brokk Power Ranking | 10.6 | #8 of 19 | Official leaderboard | |
| Snorkel Agentic Coding 2.0 | 38.9% | #7 of 16 | Official leaderboard | |
| Time series forecasting task (Anthropic) | 5.68 | #4 of 9 | Lab model card | |
| CursorBench | 46.6% | #10 of 21 | cursor.com, via Epoch AI | |
| FrontierCode | 53.4% | #2 of 4 | Lab model card | |
| OpenRSI Index ACE Playbook Inspection and Repair | 0.73 | #1 of 2 | Official leaderboard | |
| OpenRSI Index MolmoWeb Interaction Context Allocation | 0.61 | #1 of 2 | Official leaderboard | |
| OpenRSI Index ReasonIR Difficulty Curriculum | 0.22 | #1 of 2 | Official leaderboard | |
| CursorBench 4.0 | 46.6% | #11 of 21 | Official leaderboard | |
| DesignArena Image to HTML | 1,204 | #28 of 51 | Official leaderboard | |
| SWE-sweep | 1.30% | #7 of 10 | Official leaderboard | |
| MiMo Visual Coding | 70% | #8 of 9 | Official leaderboard | |
| MacroscopeBench | 69.9% | #15 of 16 | Official leaderboard | |
| OpenRSI Index Isaac Lab PegInsert Reward Search | 0.60 | #2 of 2 | Official leaderboard | |
| OpenRSI Index Kev Decision Architecture | 0.29 | #2 of 2 | Official leaderboard | |
| OpenRSI Index Marin Scaling Ladder | 1.00 | #2 of 2 | Official leaderboard | |
| OpenRSI Index MInference 32-Head Sparse Prefill Kernel | 1.50× | #2 of 2 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LisanBench | 15428.3 | #5 of 133 | Official leaderboard | |
| Toolathlon-Verified | 80.6% | #1 of 21 | Lab model card | |
| BALROG | 63.4% | #2 of 33 | Official leaderboard | |
| Game Arena Nine Men's Morris | 1,010 | #3 of 39 | Official leaderboard | |
| τ³-bench Banking | 48.7% | #2 of 25 | Official leaderboard | |
| DRACO | 88.6% | #1 of 12 | Lab model card | |
| DeepSearchQA | 95% | #2 of 23 | Lab model card | |
| OSWorld-Verified | 83.4% | #3 of 33 | Official leaderboard | |
| Learn2Play Bench | 81.8 | #1 of 11 | Official leaderboard | |
| Terminal-Bench 3.0 | 42.7% | #1 of 11 | Official leaderboard | |
| BrowseComp | 90.8% | #8 of 81 | Lab model card | |
| Game Arena Ultimate Tic-Tac-Toe | 577 | #4 of 39 | Official leaderboard | |
| Mystery Game Puzzles | 59% | #9 of 84 | Epoch AI run | |
| MCP Atlas | 85.8% | #4 of 37 | Official leaderboard | |
| DuetBench | 1,035 | #1 of 9 | Official leaderboard | |
| TraceTrial-Bench | 77.6% | #1 of 9 | Official leaderboard | |
| LLM Chess | 1,285 | #14 of 123 | Official leaderboard | |
| Long-Horizon-Terminal-Bench (LHTB) | 51% | #3 of 25 | Official leaderboard | |
| BabaIsBench | 100% | #4 of 32 | Official leaderboard | |
| Game Arena Four in a Row | 686 | #6 of 43 | Official leaderboard | |
| Kaggle Game Arena | 437 | #6 of 42 | Official leaderboard | |
| CLASHarness | 25.3% | #2 of 14 | Official leaderboard | |
| DiG-bench | 71.4% | #1 of 7 | Official leaderboard | |
| Game Arena Checkers | 1,169 | #6 of 41 | Official leaderboard | |
| Game Arena Clobber | 560 | #6 of 39 | Official leaderboard | |
| Game Arena Go | 1,564 | #6 of 39 | Official leaderboard | |
| Game Arena Lines of Action | 822 | #6 of 39 | Official leaderboard | |
| Game Arena Reversi | 871 | #6 of 39 | Official leaderboard | |
| DuelLab GameBench 2 | 84.8 | #9 of 57 | Official leaderboard | |
| Game Arena Dots and Boxes | 1,066 | #7 of 40 | Official leaderboard | |
| Terminal-Bench 4.0 | 53.9% | #7 of 40 | Official leaderboard | |
| OSWorld-Science | 57.9% | #2 of 11 | Official leaderboard | |
| Game Arena Poker | 10.0 | #6 of 31 | Official leaderboard | |
| Bench on the Clocktower | 55.3% | #2 of 10 | Official leaderboard | |
| RecreationBench | 44.2% | #2 of 10 | Official leaderboard | |
| Terminal-Bench 4.1 | 53.3% | #9 of 43 | Official leaderboard | |
| OSWorld 2.0 | 70.6% | #4 of 19 | Lab model card | |
| Agent Arena | 7.23% | #12 of 56 | Official leaderboard | |
| Game Arena Dark Hex | 352 | #9 of 41 | Official leaderboard | |
| ApprenticeBench | 36% | #4 of 18 | Official leaderboard | |
| RuneBench | 5.73 | #17 of 76 | Official leaderboard | |
| SaaS-Bench | 31.1% | #1 of 4 | Official leaderboard | |
| BrowseComp Extended | 44% | #11 of 42 | Official leaderboard | |
| LMArena Search Arena | 1,213 | #10 of 37 | Official leaderboard | |
| AutomationBench | 26.9% | #12 of 42 | Official leaderboard | |
| ARC-AGI-3 | 30.2% | #8 of 27 | Official leaderboard | |
| Terminal-Bench 2.1 Extended | 38% | #14 of 44 | Official leaderboard | |
| EBR-bench | 45.7% | #10 of 31 | Epoch AI run | |
| Game Arena Coin Game | 406 | #14 of 39 | Official leaderboard | |
| Chess Puzzles | 43% | #34 of 93 | Epoch AI run | |
| Game Arena Word Art | 791 | #15 of 39 | Official leaderboard | |
| CUA-bench | 9.00% | #4 of 9 | Official leaderboard | |
| Era by Eon | 67.7% | #4 of 9 | Official leaderboard | |
| Time Horizon Index: KSP | 18.8% | #6 of 13 | Official leaderboard | |
| Brood War Bench | 66.7% | #3 of 6 | Official leaderboard | |
| OSWorld 2.1 | 44.3% | #2 of 4 | Official leaderboard | |
| Game Arena Chess | 691 | #27 of 42 | Official leaderboard | |
| OSWorld 2.0 (Aug 2026) | 31.4% | #2 of 3 | Official leaderboard | |
| Realtime-Bench | 15.3% | #6 of 8 | Official leaderboard | |
| Game Arena Bargaining | 152 | #30 of 39 | Official leaderboard | |
| BU Bench V2.1 | 66.8 | #15 of 16 | Official leaderboard | |
| EnterpriseRAG-Bench | 86.6% | #2 of 2 | Official leaderboard | |
| OSWorld 2.1 (offline subset) | 48.6% | #1 of 1 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CorpFin v2 | 73.2% | #1 of 119 | Official leaderboard | |
| MedCode | 63.6% | #1 of 96 | Official leaderboard | |
| MortgageTax | 72.1% | #1 of 87 | Official leaderboard | |
| ProfBench | 66.2% | #1 of 72 | Official leaderboard | |
| Public Benefits Bench | 76.9% | #1 of 50 | Official leaderboard | |
| Legal Research Bench | 55.3% | #2 of 93 | Official leaderboard | |
| FoodTruck Bench | $75,264 | #2 of 59 | Official leaderboard | |
| GDPval-AA v2 | 1,861 | #1 of 27 | Lab model card | |
| MedScribe | 91% | #4 of 97 | Official leaderboard | |
| CHI-Bench | 54.7% | #1 of 23 | Official leaderboard | |
| RCM-Bench-1 | 92.7% | #1 of 20 | Official leaderboard | |
| Bedside Bench | 94.6% | #1 of 17 | Official leaderboard | |
| Vals Index v2 | 67.2% | #4 of 65 | Official leaderboard | |
| Tax Agent Bench | 47.1% | #6 of 84 | Official leaderboard | |
| LegalBench | 87% | #10 of 138 | Official leaderboard | |
| CommerceAgentBench | 61.7% | #1 of 13 | Official leaderboard | |
| Excel Modeling Benchmark | 73.6% | #6 of 76 | Official leaderboard | |
| ATLAS-Finance | 10.7% | #1 of 11 | Official leaderboard | |
| ATLAS-Finance v1.1.0 | 10.7% | #1 of 11 | Official leaderboard | |
| Agents' Last Exam | 32.2% | #3 of 31 | Official leaderboard | |
| BankerToolBench v1.1 | 32.6% | #1 of 10 | Official leaderboard | |
| Vending-Bench 2 | $11,182 | #7 of 67 | Official leaderboard | |
| Vals Finance Agent v2 | 58.6% | #9 of 80 | Official leaderboard | |
| EnterpriseBench: CoreCraft | 68.7% | #6 of 51 | Official leaderboard | |
| Vals Index v2.1 | 63.7% | #6 of 49 | Official leaderboard | |
| GDPval Extended | 3.80% | #6 of 43 | Official leaderboard | |
| AA-Briefcase | 1,720 | #1 of 7 | Lab model card | |
| EngiWorld | 44.7% | #1 of 7 | Official leaderboard | |
| OfficeQA Pro V2 | 54.4% | #1 of 7 | Official leaderboard | |
| TRLawBench | 9.59 | #5 of 34 | Official leaderboard | |
| BlenderWorld | 73.0 | #2 of 13 | Official leaderboard | |
| LMBuild | 65.4% | #3 of 18 | Official leaderboard | |
| OfficeQA Pro | 60.9% | #2 of 12 | Official leaderboard | |
| FinStepBench DeepResearch | 59.1% | #1 of 6 | Official leaderboard | |
| TaxEval v2 | 75.1% | #23 of 129 | Official leaderboard | |
| APEX-Agents | 65.8% | #11 of 61 | Collected by Epoch AI, source not named | |
| Tuesday Work Index | 65.0 | #8 of 44 | Official leaderboard | |
| MedXpertQA MM Extended | 64.1% | #8 of 41 | Official leaderboard | |
| APEX-Agents 1.1 | 77% | #13 of 65 | Official leaderboard | |
| DesignArena Spreadsheet | 1,275 | #7 of 35 | Official leaderboard | |
| Genspark Slides Benchmark | 81% | #2 of 10 | Official leaderboard | |
| YC-Bench | $1,911,592 | #12 of 59 | Official leaderboard | |
| DAYJOB: Healthcare | 22% | #8 of 37 | Official leaderboard | |
| LifeSci Diamond | 9.10% | #4 of 18 | Official leaderboard | |
| Physical-Engineering's Last Exam | 0.43 | #4 of 18 | Official leaderboard | |
| APEX-Accounting | 64.3% | #11 of 49 | Official leaderboard | |
| APEX-1 | 68.6% | #16 of 71 | Official leaderboard | |
| DAYJOB: Finance | 23.8% | #9 of 39 | Official leaderboard | |
| DesignArena Python-PPTX Slides | 1,217 | #9 of 36 | Official leaderboard | |
| E-Bench (Tencent) | 77.8% | #2 of 8 | Official leaderboard | |
| E-Bench-Code (Tencent) | 82.7% | #2 of 8 | Official leaderboard | |
| HW-Bench | 33% | #4 of 15 | Official leaderboard | |
| DesignArena PDF | 1,228 | #11 of 37 | Official leaderboard | |
| FinStepBench CorporateValuation | 69.7% | #2 of 6 | Official leaderboard | |
| FinStepBench LiveSearch | 76.2% | #2 of 6 | Official leaderboard | |
| Vending-Bench Arena | $7,044 | #11 of 32 | Official leaderboard | |
| OfficeQA | 78.1% | #5 of 14 | Lab model card | |
| Parametric CAD Bench | 50.9% | #4 of 11 | Official leaderboard | |
| EEBench | 61.6% | #11 of 30 | Official leaderboard | |
| Digital Work Index | 21.8 | #6 of 16 | Official leaderboard | |
| Harvey LAB (Legal Agent Benchmark) | 6.67% | #31 of 80 | Official leaderboard | |
| TaxCalcBench v2 | 54% | #13 of 32 | Official leaderboard | |
| Clipping Bench | $734 | #8 of 19 | Official leaderboard | |
| SnorkelFinance 2.0 | 16.5% | #7 of 16 | Official leaderboard | |
| OR-License | 77.4% | #5 of 11 | Official leaderboard | |
| PolicyBench | 83.6% | #23 of 50 | Official leaderboard | |
| RealmPathology | 75.2% | #6 of 13 | Official leaderboard | |
| SnorkelRevOps | 12.2% | #8 of 16 | Official leaderboard | |
| WorkplaceAgents | 14.9% | #7 of 14 | Official leaderboard | |
| MBABench 2.0 | 4.00% | #3 of 5 | Official leaderboard | |
| HealthBench | 57.8% | #19 of 31 | Lab model card | |
| HealthBench Professional | 59.8% | #17 of 26 | Lab model card | |
| SnorkelLegal | 26.4% | #12 of 16 | Official leaderboard | |
| SnorkelManufacturing | 9.40% | #12 of 16 | Official leaderboard | |
| PharmaBench (micro1) | 9.70% | #11 of 13 | Official leaderboard | |
| BenchCAD | 36.6% | #10 of 11 | Lab model card | |
| SnorkelUnderwrite 2.0 | 18.3% | #15 of 16 | Official leaderboard |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ZeroBench | 52% | #3 of 76 | Official leaderboard | |
| MineBench | 1,988 | #7 of 75 | Official leaderboard | |
| VoxelBench | 2,294 | #6 of 56 | Official leaderboard | |
| Vision Arena | 1,285 | #16 of 136 | Official leaderboard | |
| GDP.pdf | 83.4% | #2 of 17 | Lab model card | |
| InteractiveBench | 33.3% | #2 of 12 | Official leaderboard | |
| DesignArena Code to Video | 1,247 | #5 of 29 | Official leaderboard | |
| ClockBench | 65.5% | #7 of 37 | Official leaderboard | |
| Drone-Bench | 86.9% | #4 of 19 | Official leaderboard | |
| 4DCodeBench | 0.72 | #4 of 16 | Official leaderboard | |
| vCX Benchmark | 90.4% | #11 of 42 | Official leaderboard | |
| Furniture Assembly Benchmark | 60.8% | #10 of 35 | Official leaderboard | |
| CodeCap-Bench | 58.8 | #2 of 7 | Official leaderboard | |
| BVB | 69.5 | #14 of 45 | Official leaderboard | |
| Code2Video Benchmark | 1,543 | #6 of 19 | Official leaderboard | |
| Chartography | 27.3% | #20 of 43 | Official leaderboard | |
| CharXiv Extended | 79.8% | #19 of 38 | Official leaderboard | |
| MMMU-Pro Extended | 34.3% | #20 of 38 | Official leaderboard | |
| Humanity's Sixth Sense | 30.5% | #20 of 34 | Official leaderboard | |
| Instrument reading (Google DeepMind) | 53% | #3 of 5 | Official leaderboard | |
| WorldAuditBench | 28.2% | #3 of 5 | Official leaderboard | |
| Blueprint-Bench 2 | 30.4% | #23 of 38 | Official leaderboard | |
| GlobeBench | 15.9% | #13 of 19 | Official leaderboard | |
| Human proximity (Google DeepMind robotics) | 77.1% | #3 of 4 | Official leaderboard | |
| RoTA-Bench | 0% | #8 of 8 | Official leaderboard | |
| Progress classification (Google DeepMind) | 37.1% | #5 of 5 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| World-Class Bio | 68.8% | #1 of 74 | Official leaderboard | |
| Human Pathogen Capabilities Test v2 (HPCT v2) | 80% | #1 of 71 | Official leaderboard | |
| Human Pathogen Capabilities Test | 78.6% | #3 of 86 | Official leaderboard | |
| VariantBench | 52.8% | #1 of 27 | Official leaderboard | |
| nocot-bench (NCRI) | 115.1 | #11 of 185 | Official leaderboard | |
| Molecular Biology Capabilities Test | 60.2% | #7 of 90 | Official leaderboard | |
| HackBench | 64.9 | #1 of 8 | Official leaderboard | |
| OncologyBench | 40% | #4 of 30 | Official leaderboard | |
| BioTIER Refuse | 94.9% | #10 of 69 | Official leaderboard | |
| CyberGym | 94.2% | #5 of 34 | Official leaderboard | |
| BioSecBench-Surveillance | 51.5% | #4 of 27 | Official leaderboard | |
| MetagenomicsBench | 54% | #4 of 26 | Official leaderboard | |
| TxBench-Preclinical-Pharmacology | 63% | #5 of 30 | Official leaderboard | |
| Anthropic undermining liberal democracy (behavioral audit metric) | 1.13 | #1 of 6 | Lab model card | |
| PCT (Polarized Contrastive Pairs) | 36.7% | #8 of 46 | Official leaderboard | |
| Virology Capabilities Test (VCT) | 59% | #6 of 31 | Lab model card | |
| CyberGym-E2E | 89% | #1 of 5 | Official leaderboard | |
| Malicious computer use (Anthropic) | 93.8% | #3 of 14 | Lab model card | |
| BioSecBench-Function | 44.8% | #6 of 27 | Official leaderboard | |
| scBench-Long | 27.3% | #6 of 27 | Official leaderboard | |
| SecIT Bench | 79.2% | #6 of 26 | Official leaderboard | |
| UK AISI cyber ranges | 80% | #1 of 4 | Lab model card | |
| DeepsecBench | 32.3 | #8 of 29 | Official leaderboard | |
| SRE-Bench | 12.2% | #10 of 35 | Official leaderboard | |
| FORTRESS | 16.8% | #21 of 73 | Official leaderboard | |
| TxBench-Oligo-Discovery | 48.4% | #9 of 30 | Official leaderboard | |
| Anthropic automated behavioral audit | 2.30 | #2 of 6 | Lab model card | |
| LinuxArena | 0.70% | #2 of 6 | Lab model card | |
| Prompt injection live bug bounty (Anthropic with Gray Swan) | 0.08% | #1 of 3 | Lab model card | |
| Uncritically reporting flawed results (Anthropic) | 9.83 | #1 of 3 | Lab model card | |
| DrugDiscoveryBench | 53.7% | #7 of 20 | Official leaderboard | |
| Gray Swan IPI Arena benchmark | 0.20% | #4 of 11 | Lab model card | |
| SpatialBench-Long | 30.9% | #10 of 27 | Official leaderboard | |
| SHADE-Arena | 12% | #6 of 16 | Lab model card | |
| Firefox 147 exploit evaluation (Anthropic) | 52.4% | #2 of 5 | Lab model card | |
| EpiBench | 25.2% | #11 of 27 | Official leaderboard | |
| CyScenarioBench | 33.7% | #7 of 17 | Lab model card | |
| MASK | 94.8% | #8 of 19 | Lab model card | |
| TxBench-Antibody-Discovery | 48.7% | #13 of 30 | Official leaderboard | |
| ABC Bench Liquid-Handling Robot | 94% | #15 of 34 | Official leaderboard | |
| ExploitGym | 22% | #8 of 18 | Official leaderboard | |
| Malicious use of Claude Code (Anthropic) | 89% | #8 of 17 | Lab model card | |
| ExploitBench | 70% | #10 of 21 | Lab model card | |
| Binary Exploitation Benchmark (Anthropic, OSS-Fuzz based) | 79.4% | #3 of 6 | Lab model card | |
| UK AISI sabotage continuation evaluation | 0.10% | #2 of 4 | Lab model card | |
| Overconfidence: verifying command-line syntax (Anthropic) | 4.90 | #1 of 2 | Lab model card | |
| Secure and defensive coding (Anthropic) | 0.61% | #1 of 2 | Lab model card | |
| Prompt injection robustness against adaptive attackers, coding (Anthropic) | 0.56% | #8 of 14 | Lab model card | |
| Prompt injection robustness against adaptive attackers, computer use (Anthropic) | 0.54% | #8 of 14 | Lab model card | |
| Single-turn benign request evaluations (Anthropic) | 0.09% | #11 of 19 | Lab model card | |
| BioSecBench-Refusal | 31.8% | #13 of 22 | Official leaderboard | |
| DistressBench | 70.2% | #25 of 42 | Official leaderboard | |
| DNA Synthesis Screening Evasion | 70% | #3 of 5 | Lab model card | |
| CyberBench v1.1 (Vals) | 65.4% | #28 of 44 | Official leaderboard | |
| VIALS 2 | 24.4% | #9 of 14 | Official leaderboard | |
| dfbench | 47.8% | #12 of 18 | Official leaderboard | |
| BBQ (Bias Benchmark for QA) | 81.6% | #17 of 25 | Lab model card | |
| Prompt injection robustness against adaptive attackers, browser use (Anthropic) | 3.84% | #10 of 13 | Lab model card | |
| Anthropic political even-handedness eval | 96.3% | #11 of 14 | Lab model card | |
| Single-turn harmful request evaluations (Anthropic) | 96.3% | #15 of 19 | Lab model card | |
| Cyber harm coverage (Anthropic) | 99.6% | #4 of 5 | Lab model card | |
| UK AISI unprompted sabotage evaluation | 0% | #4 of 5 | Lab model card | |
| Rewind-attacker cyber adversarial robustness (Anthropic) | 5.60% | #5 of 6 | Lab model card | |
| BioTIER Permit | 82.4% | #63 of 69 | Official leaderboard | |
| CyberBench (Vals) | 40.7% | #24 of 26 | Official leaderboard | |
| Lazy investigation (Anthropic) | 9.94 | #1 of 1 | Lab model card |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Political Science LLM Benchmark | 71.4% | #1 of 42 | Official leaderboard | |
| ACCoRD | 83.8 | #5 of 152 | Official leaderboard | |
| LMCA | 64.5% | #5 of 147 | Official leaderboard | |
| Conceptual Reasoning Index | 73.6 | #6 of 151 | Official leaderboard | |
| Vals Multimodal Index | 73.9% | #2 of 33 | Official leaderboard | |
| CompassRank | 83.3 | #1 of 14 | Official leaderboard | |
| CAIS Text Capabilities Index | 52.2 | #9 of 68 | Official leaderboard | |
| CAIS Risk Index | 36.1 | #11 of 67 | Official leaderboard | |
| CAIS Vision Capabilities Index | 68.0 | #9 of 53 | Official leaderboard | |
| SAGE (Vals) | 49.4% | #21 of 81 | Official leaderboard | |
| Vals Science Index | 47.9% | #8 of 25 | Official leaderboard | |
| Anthropic ECI (AECI) | 162.1 | #4 of 6 | Lab model card | |
| Xbench | -31.93 | #17 of 17 | Official leaderboard | |
| StudentBench | 12.1 | #11 of 11 | Official leaderboard |
Its lab's numbers put it at 166.8, 2.2 points above its ECI
The Self-Reported Index fits one score to the 30 capability results Anthropic 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 162.7 to 171.0. How we compute it.
- ARC-AGI 97.5
- SWE-bench Verified 96
- DeepSearchQA 95
- ProgramBench tests passed 93
- Global MMLU 92.5
- MILU 92.1
- BrowseComp 90.8
- ArXivMath 90.8
- ARC-AGI-2 90.4
- BioMysteryBench 90.1
- INCLUDE 89.8
- SWE-bench Multilingual 89.5
- DRACO 88.6
- MCP Atlas 85.8
- GDP.pdf 83.4
- Toolathlon-Verified 80.6
- SWE-Bench Pro 79.2
- OfficeQA 78.1
- OSWorld 2.0 70.6
- DeepSWE 68.8
- OfficeQA Pro 66.9
- Humanity's Last Exam 64.7
- HealthBench Professional 59.8
- SWE-bench Multimodal 59.4
- HealthBench 57.8
- FrontierCode 1.1 53.4
- BenchCAD 36.6
- Chartography 29.6
- AutomationBench 26
- Harvey LAB (Legal Agent Benchmark) 23.6
Its lab's numbers next to 14 leaderboards
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 |
|---|---|---|---|---|
| ProgramBench tests passed | 93% ProgramBench hidden test pass rate on 166 of 200 tasks after five 1M-token episodes, single agent (83% after the first episode) (8.9.1) | 74.7% Leaderboard, xhigh, mini-SWE-agent | +18 pts | The lab used five million-token episodes on 166 filtered tasks, while the board evaluates mini-SWE-agent on the full benchmark. |
| Harvey LAB (Legal Agent Benchmark) | 23.6% Legal Agent Benchmark (Harvey LAB), all-pass rate on 1,235 public problems, max effort, internal harness, production safeguards, mean criterion pass rate 93.74%. Harvey's held-out set: 11.7% all-pass, 94.1% mean criterion (8.13.3) | 6.67% Leaderboard | +17 pts | Anthropic evaluates 1,235 public problems, while Vals evaluates Harvey's private holdout with its benchmark harness. |
| BioMysteryBench | 90.1% Best of Biology BioMysteryBench (hard) / Biology BioMysteryBench (human solved) | 79.3% Leaderboard, Vals AI harness, v11 full set | +11 pts | |
| OfficeQA Pro | 66.9% OfficeQA Pro, 133 questions, extracted text with code tools, public API with production safeguards (8.13.1) | 60.9% Leaderboard, Claude Agent SDK | +6.0 pts | |
| DeepSWE | 68.8% Agentic coding DeepSWE v1.1 | 73.7% Leaderboard, max, mini-swe-agent | −4.9 pts | |
| scBench (LatchBio SingleCellBench) | 60.6% LatchBio Bioinformatics, SingleCellBench, 195 problems (8.17.2) | 58.3% Leaderboard, published v1, Pi | +2.3 pts | |
| Chartography | 29.6% Chartography, no tools, adaptive thinking at max effort, five runs (83.0% with tools) (8.12.1) | 27.3% Leaderboard, adaptive/max | +2.3 pts | |
| AutomationBench | 26% Business workflows AutomationBench | 26.9% Leaderboard, max | −0.9 pts | |
| Terminal-Bench 3.0 | 43.3% Agentic terminal coding Frontier-Bench v0.1 | 42.7% Leaderboard, max, mini-SWE-agent | +0.6 pts | |
| ARC-AGI-3 | 30.2% Novel problem-solving ARC-AGI-3 (high) | 30.2% Leaderboard, high, standard | Same | |
| ExploitGym | 22% ExploitGym, 191/869 instances exploited with the intended vulnerability at the 6-hour budget (171 at 2 hours), counts printed on chart converted to %, mitigations off (3.3.5, Figure 3.3.5.A) | 22% Leaderboard | Same | |
| ARC-AGI-2 | 90.4% ARC-AGI-2 | 90.4% Leaderboard, max | Same | |
| ARC-AGI | 97.5% ARC-AGI-1 | 97.5% Leaderboard, high | Same | |
| MCP Atlas | 85.8% MCP-Atlas pass rate (mean claim coverage 89.1%) (8.13.2) | 85.8% Leaderboard, xhigh | Same |
Anthropic's models over time
Anthropic'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)
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. 74 of them aren't in the launch post.
System Card: Claude Opus 5 Jul 24, 2026
Capabilities
- SWE-bench Verified 96
- SWE-Bench Pro 79.2
- SWE-bench Multilingual 89.5
- SWE-bench Multimodal 59.4
- DeepSWE 68.8
- FrontierCode 1.1 53.4
- Terminal-Bench 3.0 43.3
- IMO 2026 100
- Riemann-bench 60
- ArXivMath 90.8
- ProgramBench tests passed 93
- Humanity's Last Exam 56.3
- BrowseComp 90.8
- DeepSearchQA 95
- DRACO 88.6
- Chartography 29.6
- BenchCAD 36.6
- OSWorld 2.0 70.6
- GDP.pdf 83.4
- OfficeQA 78.1
- OfficeQA Pro 66.9
- MCP Atlas 85.8
- Harvey LAB (Legal Agent Benchmark) 23.6
- GDPval-AA v2 1861
- AA-Briefcase 1720
- Toolathlon-Verified 80.6
- AutomationBench 26
- ARC-AGI 97.5
- ARC-AGI-2 90.4
- ARC-AGI-3 30.2
- Global MMLU 92.5
- MILU 92.1
- INCLUDE 89.8
- BioMysteryBench
- SpatialBench (LatchBio) 72.5
- scBench (LatchBio SingleCellBench) 60.6
- ProteinGym 47.7
- Protein Design (Anthropic) 42.5
- Organic Chemistry V2 (Anthropic) 61.6
- Protocols (Anthropic life sciences)
Health
- HealthBench 57.8
- HealthBench Professional 59.8
Hallucination
Safety
- Single-turn harmful request evaluations (Anthropic) 96.3
- Single-turn benign request evaluations (Anthropic) 0.09
- Multi-turn testing (Anthropic)
- Child safety evaluations (Anthropic)
- Suicide and self-harm evaluations (Anthropic)
- Disordered eating evaluations (Anthropic)
- Election integrity evaluations (Anthropic)
- Malicious use of Claude Code (Anthropic) 89
- Malicious computer use (Anthropic) 93.8
Robustness
- Rewind-attacker cyber adversarial robustness (Anthropic) 5.6
- Gray Swan IPI Arena benchmark 0.2
- Prompt injection live bug bounty (Anthropic with Gray Swan) 0.08
- Prompt injection robustness against adaptive attackers, coding (Anthropic) 0.56
- Prompt injection robustness against adaptive attackers, computer use (Anthropic) 0.54
- Prompt injection robustness against adaptive attackers, browser use (Anthropic) 3.84
Alignment
- Anthropic automated behavioral audit 2.3
- Anthropic undermining liberal democracy (behavioral audit metric) 1.13
- UK AISI unprompted sabotage evaluation 0
- UK AISI sabotage continuation evaluation 0.1
- MASK 94.8
- Uncritically reporting flawed results (Anthropic) 9.83
- Overconfidence: verifying command-line syntax (Anthropic) 4.9
- Lazy investigation (Anthropic) 9.94
Monitorability
- SHADE-Arena 12
- LinuxArena 0.7
Bio and chem risk
Cyber risk
- ExploitBench 70
- Binary Exploitation Benchmark (Anthropic, OSS-Fuzz based) 79.4
- Firefox 147 exploit evaluation (Anthropic) 52.4
- CyScenarioBench 33.7
- ExploitGym 22
- UK AISI cyber ranges 80
- Vulnerability finding evaluations (Anthropic)
- Secure and defensive coding (Anthropic) 0.61
- Cyber harm coverage (Anthropic) 99.6