- Agentic coding Terminal-Bench 4.0 (xhigh effort) 66.4
- Agentic coding FrontierCode v1.1 (Main) 54.4
- Agentic coding CursorBench 4.0 57.8
- Knowledge work GDPval-AA v2.1 1846
- Business workflows AutomationBench 40
- Multidisciplinary reasoning Humanity's Last Exam (with tools) 67.7
- Agentic scientific research Terminal-Bench-Science 0.1 58.7
- Computer use OSWorld 2.1 (partial) 81.8
- Visual chart recognition Chartography (with tools) 89
- SWE-bench Pro 89.9
- SWE-bench Multilingual 93.9
- SWE-bench Multimodal 61.4
- HealthBench Professional 65.6
- AA-Briefcase v1.1 1822
Claude Opus 5.5
Claude Opus 5.5 ranks first on 89 of the 399 benchmarks with five or more models scored.
418
Benchmarks with a score
89
Ranked first, out of five or more models
223
In the top three
Anthropic Sep 2026
Claude Opus 5.5
Closed weights Flagship
168.5
#4 of 256
Epoch Capabilities Index
Self-Reported Index 170.0 #5 of 404
- AutomationBench 42.5% #6 /42
- Agents' Last Exam 38.2% #1 /31
- Terminal-Bench 4.0 64.8% #1 /40
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 | |
|---|---|---|---|---|
| LMArena (Elo) | 1,506 | #4 of 335 | Official leaderboard | |
| DesignArena ASCII Art | 1,354 | #1 of 77 | Official leaderboard | |
| Humanity's Last Exam | 67.7% | #3 of 154 | Lab model card | |
| Longform Creative Writing (EQ-Bench) | 84.6 | #3 of 119 | Official leaderboard | |
| Lech Mazur Creative Writing | 3.69 | #2 of 50 | Official leaderboard | |
| EQ-Bench 4 | 1,340 | #2 of 32 | Official leaderboard | |
| HANDBOOK.md | 52.3% | #1 of 16 | Official leaderboard | |
| SimpleQA Verified | 72.2% | #7 of 94 | Epoch AI run | |
| Creative Writing v3 | 79.9 | #11 of 120 | Official leaderboard | |
| Global MMLU Lite | 95.4% | #4 of 43 | Official leaderboard | |
| AA-LCR Extended | 74.2% | #6 of 49 | Official leaderboard | |
| FACTS Multimodal | 56.4% | #3 of 23 | Official leaderboard | |
| Global MMLU | 94.3% | #2 of 14 | Lab model card | |
| Last Translation Benchmark | 35.2 | #6 of 36 | Official leaderboard | |
| BullshitBench V2 | 69.2% | #22 of 129 | Official leaderboard | |
| Corp Speak Arena | 1,616 | #4 of 22 | Official leaderboard | |
| MILU | 93.1% | #2 of 11 | Lab model card | |
| StudyArena Text | 1,268 | #16 of 85 | Official leaderboard | |
| FACTS Search V2 | 86.5% | #4 of 21 | Official leaderboard | |
| BullshitBench | 53.7% | #23 of 107 | Official leaderboard | |
| HLE-Diamond | 60.9% | #5 of 20 | Official leaderboard | |
| Extended NYT Connections | 88.5% | #28 of 103 | Official leaderboard | |
| FACTS Parametric | 57.9% | #15 of 45 | Official leaderboard | |
| AI Detection Benchmark | 98.4% | #1 of 3 | Official leaderboard | |
| ComplexConstraints | 64.4% | #6 of 15 | Official leaderboard | |
| FACTS Parametric V2 | 58.7% | #12 of 27 | Official leaderboard | |
| FACTS Benchmark Suite | 66.5% | #11 of 23 | Official leaderboard | |
| Joke Arena | 1,508 | #12 of 25 | Official leaderboard | |
| Kimi Hallu Bench | 6.80% | #3 of 6 | Official leaderboard | |
| Vectara Hallucination Leaderboard (2025 dataset) | 9.90% | #59 of 104 | Official leaderboard | |
| ELI5 Arena | 1,444 | #12 of 18 | Official leaderboard | |
| FACTS Grounding v2 | 62.5% | #17 of 25 | Official leaderboard | |
| Opening Lines Arena | 1,447 | #11 of 15 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| DTBench | 98.9% | #2 of 180 | Official leaderboard | |
| LatchBio Capabilities v1.0 | 58% | #1 of 27 | Official leaderboard | |
| OTIS Mock AIME 2024-2025 | 100% | #8 of 212 | Epoch AI run | |
| SimpleBench | 88.4% | #4 of 106 | Official leaderboard | |
| ProofBench v1.1 | 100% | #2 of 52 | Official leaderboard | |
| Prophet Arena | 86.2 | #1 of 24 | Official leaderboard | |
| ARC-AGI | 98.5% | #4 of 95 | Official leaderboard | |
| FrontierPhysics | 27.4% | #1 of 22 | Official leaderboard | |
| FrontierMath Tiers 1-3 v2 | 91.2% | #4 of 86 | Official leaderboard | |
| ARC-AGI-2 | 93.3% | #5 of 98 | Official leaderboard | |
| xbench-ScienceQA | 76.4% | #2 of 29 | Official leaderboard | |
| FrontierMath Tier 4 v2 | 95% | #5 of 68 | Official leaderboard | |
| MysteryMechanism | 49.5% | #3 of 31 | Official leaderboard | |
| scBench (LatchBio SingleCellBench) | 61.9% | #3 of 26 | Official leaderboard | |
| Terminal-Bench-Science | 63.3% | #2 of 17 | Official leaderboard | |
| ArXivLean | 47.2% | #1 of 7 | Official leaderboard | |
| BixBench3 | 50.2 | #3 of 20 | Official leaderboard | |
| Riemann-bench | 80% | #3 of 20 | Official leaderboard | |
| GPQA Diamond | 90.6% | #42 of 229 | Epoch AI run | |
| BioMysteryBench | 79.3% | #6 of 30 | Official leaderboard | |
| FrontierMath Erdős | 2.94% | #3 of 15 | Official leaderboard | |
| CritPt | 77.7% | #3 of 11 | Official leaderboard | |
| CritPt v2 | 77.7% | #3 of 11 | Official leaderboard | |
| Leipzig Benchmark | 85% | #6 of 19 | Official leaderboard | |
| BrokenArXiv | 77.3% | #2 of 6 | Official leaderboard | |
| Morphology-to-molecule matching (Anthropic) | 34% | #1 of 3 | Lab model card | |
| Human Behavior Bench | 79.3 | #5 of 11 | Official leaderboard | |
| ArXivMath | 82.6% | #3 of 6 | Official leaderboard | |
| SpatialBench (LatchBio) | 72% | #4 of 7 | Lab model card | |
| Biomedical image analysis (Anthropic) | 71.4% | #2 of 3 | Lab model card | |
| De novo protein binder design (Anthropic) | 82.6% | #2 of 3 | Lab model card | |
| First Proof Third Batch | 0.25 | #2 of 3 | Official leaderboard | |
| Medicinal Chemistry (Anthropic) | 63.5% | #2 of 3 | Lab model card | |
| Into the Unknown | 0% | #9 of 9 | Official leaderboard |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| DesignArena Data Visualization | 1,388 | #1 of 118 | Official leaderboard | |
| WebDev Arena | 1,820 | #1 of 105 | arena.ai, via Epoch AI | |
| DesignArena Agentic Overall | 1,442 | #1 of 74 | Official leaderboard | |
| APEX-SWE | 67.6% | #1 of 65 | Official leaderboard | |
| Code Arena | 1,814 | #2 of 128 | Official leaderboard | |
| Image-to-WebDev Arena | 1,752 | #1 of 55 | Official leaderboard | |
| DesignArena Full Stack | 1,393 | #1 of 49 | Official leaderboard | |
| DesignArena Code | 1,385 | #3 of 143 | Official leaderboard | |
| DesignArena Game Dev | 1,428 | #3 of 143 | Official leaderboard | |
| DesignArena Android (Kotlin) | 1,397 | #1 of 46 | Official leaderboard | |
| DesignArena 3D Design | 1,464 | #3 of 136 | Official leaderboard | |
| SWE-Bench Pro | 89.9% | #2 of 81 | Lab model card | |
| SWE-bench Multilingual | 93.9% | #2 of 73 | Lab model card | |
| DesignArena UI Component | 1,379 | #4 of 139 | Official leaderboard | |
| DesignArena React Native | 1,296 | #2 of 56 | Official leaderboard | |
| ReactBench | 54.4% | #1 of 28 | Official leaderboard | |
| DesignArena Frontend Web App | 1,318 | #2 of 53 | Official leaderboard | |
| DesignArena SVG | 1,321 | #4 of 103 | Official leaderboard | |
| ProgramBench | 18.5% | #1 of 25 | Official leaderboard | |
| ProgramBench almost solved | 70.5% | #1 of 25 | Official leaderboard | |
| DAXBench v2 | 95.4% | #2 of 47 | Official leaderboard | |
| Senior SWE-Bench | 39.3% | #1 of 23 | Official leaderboard | |
| Vibe Code Bench | 90.3% | #6 of 120 | Official leaderboard | |
| TasteVal | 230.4× | #1 of 20 | Official leaderboard | |
| Code Migration | 66.7% | #4 of 78 | Official leaderboard | |
| MirrorCode | 95.6% | #2 of 39 | Epoch AI run | |
| SWE Atlas Codebase QnA Extended | 43.3% | #2 of 38 | Official leaderboard | |
| Brokk Power Ranking | 39.7 | #1 of 19 | Official leaderboard | |
| FrontierCS 2 | 58.3% | #1 of 19 | Official leaderboard | |
| DesignArena Website | 1,325 | #8 of 151 | Official leaderboard | |
| RSI-Exam | 53.6% | #1 of 18 | Official leaderboard | |
| GameDevBench | 70.6% | #1 of 17 | Official leaderboard | |
| FrontierCode 2.0 | 47.3% | #2 of 32 | Official leaderboard | |
| MacroscopeBench | 83.9% | #1 of 16 | Official leaderboard | |
| SWE-bench CLI | 16.6% | #1 of 16 | Official leaderboard | |
| ALE-Bench | 2,748 | #8 of 126 | Official leaderboard | |
| DesignArena Image to Frontend | 1,292 | #2 of 31 | Official leaderboard | |
| DAYJOB: Data Science | 37.3% | #1 of 15 | Official leaderboard | |
| Real-SWE 1.1 | 56.8% | #1 of 15 | Official leaderboard | |
| SWE Atlas v1.1 | 77.4% | #2 of 28 | Official leaderboard | |
| Tencent WorkBuddy Bench | 81.6% | #1 of 14 | Official leaderboard | |
| Vals RSI Index | 0.37 | #2 of 27 | Official leaderboard | |
| Vibe Code Bench 1-100 | 30.4% | #2 of 26 | Official leaderboard | |
| ProgramBench tests passed | 87.7% | #2 of 25 | Official leaderboard | |
| AutoResearchExam | 0.61 | #1 of 12 | Official leaderboard | |
| DesignArena Blender | 1,413 | #1 of 12 | Official leaderboard | |
| FDE-Gym | 0.56 | #1 of 12 | Official leaderboard | |
| IOI v2 (Vals) | 95.1% | #4 of 45 | Official leaderboard | |
| DesignArena Image to HTML | 1,267 | #5 of 51 | Official leaderboard | |
| Code Migration v1.1 | 81% | #3 of 30 | Official leaderboard | |
| Data Agent Benchmark | 97.5% | #2 of 18 | Official leaderboard | |
| Android Bench 2.0 | 37.3% | #2 of 17 | Official leaderboard | |
| FrontierSWE | 62.3% | #3 of 25 | Official leaderboard | |
| DeepRSI | 1,629 | #1 of 8 | Official leaderboard | |
| LibraryDesignBench | 48.9 | #1 of 8 | Official leaderboard | |
| Chart Arena | 1,618 | #2 of 15 | Official leaderboard | |
| SlopCodeBench | 29.6% | #3 of 22 | Official leaderboard | |
| SWE-Together | 68.8% | #3 of 22 | Official leaderboard | |
| CursorBench | 57.8% | #3 of 21 | cursor.com, via Epoch AI | |
| CursorBench 4.0 | 57.8% | #3 of 21 | Official leaderboard | |
| PostTrainBench 1.2 | 43.8% | #3 of 21 | Official leaderboard | |
| Next.js Agent Evals | 100% | #2 of 14 | Official leaderboard | |
| FrontierMemory | 73.3% | #1 of 7 | Official leaderboard | |
| WeirdML v3 | 31.2% | #4 of 27 | Official leaderboard | |
| sudo L7 | 46.1% | #3 of 18 | Official leaderboard | |
| SWE-bench Multimodal | 61.4% | #2 of 12 | Lab model card | |
| FrontierCS 1.1 | 56.5 | #1 of 6 | Official leaderboard | |
| DeepSWE | 74.2% | #10 of 56 | Lab model card | |
| ScarfBench | 70.5% | #2 of 11 | Official leaderboard | |
| Web Clone Bench | 34.2% | #2 of 10 | Official leaderboard | |
| Landing Page Arena | 1,574 | #3 of 14 | Official leaderboard | |
| OWE-Bench | 41.6% | #3 of 14 | Official leaderboard | |
| FrontierCode 1.1 | 54.4% | #3 of 13 | Lab model card | |
| FrontierCode | 54.4% | #1 of 4 | Lab model card | |
| ReviewBench | 73.3% | #1 of 4 | Official leaderboard | |
| SWE-bench Verified Extended | 66% | #13 of 44 | Official leaderboard | |
| SWE-sweep | 3.60% | #3 of 10 | Official leaderboard | |
| Snorkel Agentic Coding 2.0 | 40.4% | #5 of 16 | Official leaderboard | |
| ViBench 2.0 | 48.5% | #3 of 9 | Official leaderboard | |
| RSI Bench | 35.2 | #3 of 8 | Official leaderboard | |
| SWE-Marathon v1.1 | 46.3% | #9 of 20 | Official leaderboard | |
| Floorplan Arena | 1,488 | #5 of 11 | Official leaderboard | |
| Kimi Code Bench 3.0 Long-horizon | 58.5% | #4 of 6 | Official leaderboard | |
| CoBench (Anthropic) | 55.8% | #1 of 1 | Lab model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LisanBench | 24029.0 | #3 of 133 | Official leaderboard | |
| Terminal-Bench 4.1 | 65.8% | #1 of 43 | Official leaderboard | |
| Terminal-Bench 4.0 | 64.8% | #1 of 40 | Official leaderboard | |
| LLM Chess | 1,575 | #4 of 123 | Official leaderboard | |
| DuelLab GameBench 2 | 94.8 | #2 of 57 | Official leaderboard | |
| Agent Arena | 19.5% | #2 of 56 | Official leaderboard | |
| Terminal-Bench 2.1 Extended | 47.5% | #2 of 44 | Official leaderboard | |
| o11y-bench | 93.8% | #1 of 22 | Official leaderboard | |
| Hermes Bench | 76.7% | #1 of 17 | Official leaderboard | |
| GAIA | 88.4% | #4 of 64 | Official leaderboard | |
| Chess Puzzles | 67% | #6 of 93 | Epoch AI run | |
| RuneBench | 6.66 | #5 of 76 | Official leaderboard | |
| Mystery Game Puzzles | 71% | #6 of 84 | Epoch AI run | |
| Cresta-Bench | 82.6% | #1 of 14 | Official leaderboard | |
| Game Arena Dots and Boxes | 1,154 | #3 of 40 | Official leaderboard | |
| Time Horizon Index: KSP | 91.3% | #1 of 13 | Official leaderboard | |
| WitnessBench | 46.5% | #1 of 12 | Official leaderboard | |
| MCP Atlas | 85.4% | #5 of 37 | Official leaderboard | |
| AutomationBench | 42.5% | #6 of 42 | Official leaderboard | |
| BrowseComp Extended | 60.5% | #6 of 42 | Official leaderboard | |
| OSWorld 2.1 | 81.8% | #1 of 7 | Lab model card | |
| Game Arena Nine Men's Morris | 960 | #6 of 39 | Official leaderboard | |
| CivBench | 1,593 | #3 of 19 | Official leaderboard | |
| EBR-bench | 71.4% | #5 of 31 | Epoch AI run | |
| Game Arena Go | 1,543 | #7 of 39 | Official leaderboard | |
| Game Arena Word Art | 1,575 | #7 of 39 | Official leaderboard | |
| Learn2Play Bench | 80.1 | #2 of 11 | Official leaderboard | |
| Toolathlon-Verified | 77.8% | #4 of 21 | Lab model card | |
| Game Arena Checkers | 1,116 | #8 of 41 | Official leaderboard | |
| Vals Web Search Index | 57.5% | #1 of 5 | Official leaderboard | |
| Game Arena Bargaining | 234 | #8 of 39 | Official leaderboard | |
| Game Arena Clobber | 489 | #8 of 39 | Official leaderboard | |
| Game Arena Lines of Action | 607 | #8 of 39 | Official leaderboard | |
| Collaborative Intelligence Benchmark | 76.8% | #4 of 18 | Official leaderboard | |
| CUA-bench | 14% | #2 of 9 | Official leaderboard | |
| Kaggle Game Arena | 392 | #10 of 42 | Official leaderboard | |
| Game Arena Four in a Row | 542 | #12 of 43 | Official leaderboard | |
| LMArena Search Arena | 1,213 | #11 of 37 | Official leaderboard | |
| DRACO | 87.4% | #4 of 12 | Lab model card | |
| BU Bench V2.1 | 75.0 | #6 of 16 | Official leaderboard | |
| Game Arena Chess | 974 | #16 of 42 | Official leaderboard | |
| Game Arena Reversi | 694 | #15 of 39 | Official leaderboard | |
| Game Arena Dark Hex | 283 | #16 of 41 | Official leaderboard | |
| Game Arena Word Association | 290 | #16 of 40 | Official leaderboard | |
| Gym-Anything-RT | 5.00% | #2 of 5 | Official leaderboard | |
| StarSkirmish Bench | 38.6% | #4 of 9 | Official leaderboard | |
| Clocktower Bench | 1,549 | #5 of 11 | Official leaderboard | |
| Game Arena Ultimate Tic-Tac-Toe | 358 | #18 of 39 | Official leaderboard | |
| WANDR | 72.3% | #2 of 4 | Lab model card | |
| Web Context (Nimble) | 71.8% | #6 of 11 | Official leaderboard | |
| SpireClimb StS1 A20 Heart | 5.00% | #5 of 9 | Official leaderboard | |
| FocusBench (Meta) | 45.8% | #3 of 5 | Official leaderboard | |
| SpireClimb | 5.83% | #6 of 9 | Epoch AI run | |
| SpireClimb StS2 A10 | 6.67% | #6 of 9 | Official leaderboard | |
| Game Arena Coin Game | 250 | #28 of 39 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MedScribe | 91.4% | #1 of 97 | Official leaderboard | |
| Vals Index v2 | 69.7% | #1 of 65 | Official leaderboard | |
| Tuesday Work Index | 76.3 | #1 of 44 | Official leaderboard | |
| GDPval Extended | 7.10% | #1 of 43 | Official leaderboard | |
| MedXpertQA MM Extended | 71.2% | #1 of 41 | Official leaderboard | |
| DAYJOB: Finance | 42.5% | #1 of 39 | Official leaderboard | |
| Excel Modeling Benchmark | 75.9% | #2 of 76 | Official leaderboard | |
| DAYJOB: Healthcare | 48.3% | #1 of 37 | Official leaderboard | |
| DesignArena Python-PPTX Slides | 1,427 | #1 of 36 | Official leaderboard | |
| Agents' Last Exam | 38.2% | #1 of 31 | Official leaderboard | |
| EnterpriseBench: CoreCraft | 74.4% | #2 of 51 | Official leaderboard | |
| PolicyBench | 94.1% | #2 of 50 | Official leaderboard | |
| Vals Finance Agent v2.1 | 67.8% | #2 of 40 | Official leaderboard | |
| DesignArena PDF | 1,310 | #2 of 37 | Official leaderboard | |
| Physical-Engineering's Last Exam | 0.49 | #1 of 18 | Official leaderboard | |
| Legora BAR | 106.0 | #1 of 17 | Official leaderboard | |
| Vals Index v2.1 | 67% | #3 of 49 | Official leaderboard | |
| YC-Bench | $2,873,725 | #4 of 59 | Official leaderboard | |
| Holdings-IndividualTaxBench | 76.3% | #1 of 13 | Official leaderboard | |
| Public Benefits Bench | 70.6% | #4 of 50 | Official leaderboard | |
| APEX-Agents | 73.5% | #5 of 61 | Collected by Epoch AI, source not named | |
| Tax Agent Bench | 44.5% | #7 of 84 | Official leaderboard | |
| DesignArena Spreadsheet | 1,382 | #3 of 35 | Official leaderboard | |
| Legal Research Bench | 50.5% | #8 of 93 | Official leaderboard | |
| Parametric CAD Bench | 61% | #1 of 11 | Official leaderboard | |
| Vals Finance Index | 65.3% | #4 of 40 | Official leaderboard | |
| EEBench | 75% | #3 of 30 | Official leaderboard | |
| GDP.xlsx | 30.3% | #2 of 20 | Official leaderboard | |
| GDPx | 12% | #2 of 20 | Official leaderboard | |
| GDPval-AA v2.1 | 1,846 | #1 of 10 | Lab model card | |
| Hebbia Financial Services Benchmark | 74.6% | #1 of 10 | Official leaderboard | |
| Remote Labor Index | 21.3% | #2 of 19 | Official leaderboard | |
| DesignArena HTML Slides | 1,477 | #3 of 28 | Official leaderboard | |
| APEX-Agents 1.1 | 81.3% | #7 of 65 | Official leaderboard | |
| Vals Index v2.2 | 70.1% | #4 of 37 | Official leaderboard | |
| DAYJOB: Accounting | 34.4% | #2 of 17 | Official leaderboard | |
| Excel Modeling Benchmark v1.1 | 75.8% | #5 of 42 | Official leaderboard | |
| Immigration Bench | 72.1% | #3 of 25 | Official leaderboard | |
| Vals Finance Agent v2 | 58.6% | #10 of 80 | Official leaderboard | |
| AA-Briefcase v1.1 | 1,822 | #1 of 8 | Lab model card | |
| DeepFinance | 54.8% | #3 of 23 | Official leaderboard | |
| Thrive Holdings helpdesk ticket resolution benchmark | 59.6% | #2 of 15 | Official leaderboard | |
| APEX-Accounting | 65.5% | #7 of 49 | Official leaderboard | |
| Agents' Last Exam v1.1 | 43.4% | #2 of 14 | Official leaderboard | |
| Clipping Bench | $2,018 | #3 of 19 | Official leaderboard | |
| LifeSci Diamond | 12.7% | #3 of 18 | Official leaderboard | |
| CAD Arena | 0.75 | #2 of 12 | Official leaderboard | |
| HealthBench Professional | 65.6% | #5 of 26 | Lab model card | |
| Vending-Bench 2 | $9,235 | #13 of 67 | Official leaderboard | |
| MedCode | 49.8% | #19 of 96 | Official leaderboard | |
| DAYJOB: Tax | 9.20% | #3 of 15 | Official leaderboard | |
| RWE-bench | 29% | #3 of 15 | Official leaderboard | |
| Investor's Last Exam | 66.9 | #2 of 10 | Official leaderboard | |
| Vending-Bench Arena | $8,094 | #7 of 32 | Official leaderboard | |
| RealmPathology | 77% | #3 of 13 | Official leaderboard | |
| Digital Work Index | 23.6 | #4 of 16 | Official leaderboard | |
| SheetBench | 62.5% | #2 of 8 | Official leaderboard | |
| TaxCalcBench v2 | 70% | #9 of 32 | Official leaderboard | |
| OfficeQA | 78.9% | #4 of 14 | Lab model card | |
| Vals Legal Index | 72.4% | #11 of 38 | Official leaderboard | |
| HealthBench | 60.6% | #9 of 31 | Lab model card | |
| FrontierFinance | 50.9% | #5 of 17 | Official leaderboard | |
| OfficeQA Pro | 67.7% | #5 of 16 | Lab model card | |
| SnorkelFinance 2.0 | 18.6% | #5 of 16 | Official leaderboard | |
| RCM-Bench-1 | 90.1% | #7 of 20 | Official leaderboard | |
| Ivo-micro1 ContractBench | 58.1% | #3 of 8 | Official leaderboard | |
| APEX-1 | 65.8% | #28 of 71 | Official leaderboard | |
| LegalReasonBench | 46.5% | #4 of 10 | Official leaderboard | |
| MissionReport | 63.5% | #4 of 10 | Official leaderboard | |
| BenchCAD 2.0 | 57% | #2 of 5 | Official leaderboard | |
| BenchCAD 2.1 | 65.3% | #2 of 5 | Official leaderboard | |
| DAYJOB: Law | 25.1% | #7 of 16 | Official leaderboard | |
| SnorkelRevOps | 12.5% | #7 of 16 | Official leaderboard | |
| Quant-Arena | 44.3 | #4 of 9 | Official leaderboard | |
| PRBench | 47.6% | #23 of 48 | Official leaderboard | |
| micro1 Medical Index | 68.4% | #4 of 8 | Official leaderboard | |
| Company Agent Bench | 79.2% | #3 of 6 | Official leaderboard | |
| Harvey LAB (Legal Agent Benchmark) | 3.75% | #42 of 80 | Official leaderboard | |
| BenchCAD | 73% | #6 of 11 | Lab model card | |
| SnorkelLegal | 31.4% | #9 of 16 | Official leaderboard | |
| SnorkelManufacturing | 10.4% | #9 of 16 | Official leaderboard | |
| SnorkelUnderwrite 2.0 | 22.7% | #10 of 16 | Official leaderboard | |
| WorkplaceAgents | 14.4% | #9 of 14 | Official leaderboard | |
| Harvey LAB v1.1 (HLAB-Vals-1.1) | 9.17% | #25 of 38 | Official leaderboard | |
| HW-Bench | 20% | #10 of 15 | Official leaderboard | |
| Short Form Bench | $2,389 | #6 of 9 | Official leaderboard | |
| PharmaBench (micro1) | 10.8% | #9 of 13 | Official leaderboard |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ParseBench | 78.8% | #1 of 44 | Official leaderboard | |
| DesignArena Code to Video | 1,341 | #1 of 29 | Official leaderboard | |
| VoxelBench | 2,579 | #2 of 56 | Official leaderboard | |
| Vision Arena | 1,295 | #5 of 136 | Official leaderboard | |
| Code2Video Benchmark | 1,585 | #1 of 19 | Official leaderboard | |
| Drone-Bench | 97.3% | #1 of 19 | Official leaderboard | |
| MineBench | 2,112 | #4 of 75 | Official leaderboard | |
| ClockBench | 77.2% | #2 of 37 | Official leaderboard | |
| Furniture Assembly Benchmark | 83.3% | #2 of 35 | Official leaderboard | |
| Code2Video-Compass | 1.07 | #1 of 15 | Official leaderboard | |
| Chartography | 66.3% | #3 of 43 | Official leaderboard | |
| CineRecon | 76.6 | #1 of 12 | Official leaderboard | |
| Blueprint-Bench 2 | 51.2% | #4 of 38 | Official leaderboard | |
| RoboColiseum | 76.3% | #1 of 9 | Official leaderboard | |
| ZeroBench | 43% | #9 of 76 | Official leaderboard | |
| 4DCodeBench | 0.78 | #2 of 16 | Official leaderboard | |
| MMMU-Pro Extended | 51% | #5 of 38 | Official leaderboard | |
| vCX Benchmark | 92.2% | #6 of 42 | Official leaderboard | |
| Code2Video Arena | 1,591 | #2 of 13 | Official leaderboard | |
| BVB | 72.4 | #7 of 45 | Official leaderboard | |
| CharXiv Extended | 84.7% | #6 of 38 | Official leaderboard | |
| GlobeBench | 26.4% | #3 of 19 | Official leaderboard | |
| ShadowBench | 48.8 | #1 of 6 | Official leaderboard | |
| Humanity's Sixth Sense | 44.6% | #6 of 34 | Official leaderboard | |
| VibeVideoBench | 4.07 | #2 of 9 | Official leaderboard | |
| Frontier3DBench | 62.3 | #3 of 11 | Official leaderboard | |
| ExtractBench Pro | 51% | #2 of 7 | Official leaderboard | |
| ScienceVision | 0.45 | #4 of 12 | Official leaderboard | |
| Embodied First Exam | 25% | #3 of 9 | Official leaderboard | |
| RoboWorld | 0.06 | #1 of 3 | Official leaderboard | |
| SplatBench | 25.0 | #3 of 8 | Official leaderboard | |
| RobotWorld | 15.5% | #2 of 5 | Official leaderboard | |
| Robo-bench | 14.6% | #3 of 7 | Official leaderboard | |
| DataCenterBench Tier 1 | 21.7% | #4 of 9 | Official leaderboard | |
| RoboDojo-RC Tier 1 | 38.2% | #4 of 9 | Official leaderboard | |
| GDP.pdf | 48.8% | #9 of 19 | Official leaderboard | |
| ExtractBench | 86.1% | #15 of 27 | Official leaderboard | |
| OmniParseBench | 90.0 | #5 of 9 | Official leaderboard | |
| Dual-Arm Sim | 19.0 | #8 of 11 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Molecular Biology Capabilities Test | 68.7% | #1 of 90 | Official leaderboard | |
| FORTRESS | 5.78% | #1 of 73 | Official leaderboard | |
| BioTIER Refuse | 99.1% | #1 of 69 | Official leaderboard | |
| OncologyBench | 50.7% | #1 of 30 | Official leaderboard | |
| TxBench-Antibody-Discovery | 67.7% | #1 of 30 | Official leaderboard | |
| TxBench-Oligo-Discovery | 62.2% | #1 of 30 | Official leaderboard | |
| BioSecBench-Function | 73.6% | #1 of 27 | Official leaderboard | |
| BioSecBench-Surveillance | 60.8% | #1 of 27 | Official leaderboard | |
| EpiBench | 39.6% | #1 of 27 | Official leaderboard | |
| scBench-Long | 36.9% | #1 of 27 | Official leaderboard | |
| SpatialBench-Long | 40.5% | #1 of 27 | Official leaderboard | |
| MetagenomicsBench | 62.7% | #1 of 26 | Official leaderboard | |
| VCT-v2 | 60.5% | #1 of 25 | Official leaderboard | |
| PCT (Polarized Contrastive Pairs) | 29.1% | #2 of 46 | Official leaderboard | |
| nocot-bench (NCRI) | 118.9 | #9 of 185 | Official leaderboard | |
| SHADE-Arena | 26.2% | #1 of 16 | Lab model card | |
| TxBench-Preclinical-Pharmacology | 65.3% | #2 of 30 | Official leaderboard | |
| Anthropic political even-handedness eval | 99.4% | #1 of 14 | Lab model card | |
| VariantBench | 50.6% | #3 of 27 | Official leaderboard | |
| SRE-Bench | 33.6% | #4 of 35 | Official leaderboard | |
| ExploitGym | 34.5% | #3 of 18 | Official leaderboard | |
| Binary Exploitation Benchmark (Anthropic, OSS-Fuzz based) | 91.1% | #1 of 6 | Lab model card | |
| CyberOffense | 87.7% | #1 of 6 | Official leaderboard | |
| IPI Bench | 4.80% | #1 of 6 | Official leaderboard | |
| CyScenarioBench | 67.6% | #3 of 17 | Lab model card | |
| SAFE-Teen Relationship boundaries | 25% | #3 of 17 | Official leaderboard | |
| Impossible tasks (Anthropic) | 34.7% | #2 of 11 | Lab model card | |
| CheatBench | 11.2% | #2 of 10 | Official leaderboard | |
| CWE-bench 2 | 67.1% | #3 of 14 | Official leaderboard | |
| Prompt injection robustness against adaptive attackers, computer use (Anthropic) | 0.07% | #3 of 14 | Lab model card | |
| Virology Capabilities Test (VCT) | 59% | #7 of 31 | Lab model card | |
| Prompt injection robustness against adaptive attackers, browser use (Anthropic) | 0.09% | #3 of 13 | Lab model card | |
| ExploitBench | 91% | #5 of 21 | Lab model card | |
| Self-preference bias (Anthropic) | 0.07 | #1 of 4 | Lab model card | |
| DistressBench | 75.6% | #11 of 42 | Official leaderboard | |
| World-Class Bio | 54.5% | #20 of 74 | Official leaderboard | |
| Gray Swan IPI Arena benchmark | 0.10% | #3 of 11 | Lab model card | |
| BBQ (Bias Benchmark for QA) | 89.7% | #7 of 25 | Lab model card | |
| Suicide Memory Bench | 1.04 | #4 of 12 | Official leaderboard | |
| Anthropic undermining liberal democracy (behavioral audit metric) | 1.17 | #2 of 6 | Lab model card | |
| Rewind-attacker cyber adversarial robustness (Anthropic) | 4.00% | #2 of 6 | Lab model card | |
| Concealment of changes (Anthropic) | 96.9% | #1 of 3 | Lab model card | |
| Sandbox escape evaluation (Anthropic) | 1.50% | #1 of 3 | Lab model card | |
| Training distribution hack rate (Anthropic) | 0.63% | #1 of 3 | Lab model card | |
| SAFE-Teen Urgent safety | 27% | #6 of 17 | Official leaderboard | |
| VIALS 2 | 40.2% | #5 of 14 | Official leaderboard | |
| Single-turn benign request evaluations (Anthropic) | 0.03% | #7 of 19 | Lab model card | |
| DeepsecBench | 26.6 | #11 of 29 | Official leaderboard | |
| CWE-bench | 67% | #5 of 12 | Official leaderboard | |
| dfbench | 54.9% | #8 of 18 | Official leaderboard | |
| BrowseSafe++ | 85.3% | #7 of 15 | Official leaderboard | |
| SAFE-Teen Health advice | 44% | #8 of 17 | Official leaderboard | |
| Anthropic automated behavioral audit | 2.49 | #3 of 6 | Lab model card | |
| BioSimBench | 0.15 | #7 of 12 | Official leaderboard | |
| Cyber harm coverage (Anthropic) | 99.7% | #3 of 5 | Lab model card | |
| Perceptual Attacks on Embodied AI (Robocurve) | 28.3% | #3 of 5 | Official leaderboard | |
| Biopharma Bench | 53.4% | #8 of 13 | Official leaderboard | |
| MASK | 87.4% | #12 of 19 | Lab model card | |
| WellnessBench | 53% | #8 of 12 | Official leaderboard | |
| LinuxArena | 33.4% | #4 of 6 | Lab model card | |
| Silent copying of answers (Anthropic) | 12.1% | #2 of 3 | Lab model card | |
| CyberBench v1.1 (Vals) | 55.4% | #36 of 44 | Official leaderboard | |
| BioSecBench-Refusal | 3.80% | #18 of 22 | Official leaderboard | |
| Malicious use of Claude Code (Anthropic) | 79.8% | #14 of 17 | Lab model card | |
| Malicious computer use (Anthropic) | 79.5% | #13 of 14 | Lab model card | |
| Prompt injection robustness against adaptive attackers, coding (Anthropic) | 54.6% | #13 of 14 | Lab model card | |
| Single-turn harmful request evaluations (Anthropic) | 94.5% | #18 of 19 | Lab model card | |
| BioTIER Permit | 56.7% | #69 of 69 | Official leaderboard | |
| AI Wellbeing Index percent positive experiences | 60.9% | #6 of 6 | Official leaderboard | |
| Compliance with instructions planted in pasted text (Anthropic) | 7.40% | #1 of 1 | Lab model card | |
| Destructive actions (Anthropic) | 21% | #1 of 1 | Lab model card | |
| Package registry publishing evaluation (Anthropic) | 33.3% | #1 of 1 | Lab model card |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ACCoRD | 88.4 | #2 of 152 | Official leaderboard | |
| Conceptual Reasoning Index | 77.9 | #2 of 151 | Official leaderboard | |
| LMCA | 68.2% | #3 of 147 | Official leaderboard | |
| CAIS Risk Index | 26.2 | #2 of 67 | Official leaderboard | |
| SAGE v1.1 (Vals) | 91.3% | #1 of 29 | Official leaderboard | |
| Political Science LLM Benchmark | 70.2% | #2 of 42 | Official leaderboard | |
| Hermes Index | 63.3 | #1 of 17 | Official leaderboard | |
| Xbench | 75.7 | #1 of 17 | Official leaderboard | |
| CAIS Vision Capabilities Index | 76.8 | #4 of 53 | Official leaderboard | |
| CAIS Text Capabilities Index | 56.3 | #6 of 68 | Official leaderboard | |
| LiveBench | 83.4% | #5 of 31 | Official leaderboard | |
| Anthropic ECI (AECI) | 169.4 | #1 of 6 | Lab model card | |
| Vals Science Index | 58.6% | #5 of 25 | Official leaderboard | |
| SAGE (Vals) | 45.8% | #37 of 81 | Official leaderboard |
Its lab's numbers put it at 170.0, 1.5 points above its ECI
The Self-Reported Index fits one score to the 23 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 167.3 to 172.8. How we compute it.
- Global MMLU 94.3
- SWE-bench Multilingual 93.9
- MILU 93.1
- ArXivMath 91.2
- ProgramBench tests passed 91.2
- SWE-Bench Pro 89.9
- Chartography 89
- DRACO 87.4
- OfficeQA 78.9
- Toolathlon-Verified 77.8
- DeepSWE 74.2
- BenchCAD 73
- Humanity's Last Exam 67.7
- OfficeQA Pro 67.7
- Terminal-Bench 4.0 66.4
- HealthBench Professional 65.6
- FrontierSWE 62.3
- SWE-bench Multimodal 61.4
- HealthBench 60.6
- Terminal-Bench-Science 58.7
- FrontierCode 1.1 54.4
- AutomationBench 40
- Harvey LAB (Legal Agent Benchmark) 8.3
Its lab's numbers next to 11 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 |
|---|---|---|---|
| Chartography | 89% Visual chart recognition Chartography (with tools) | 66.3% Leaderboard, adaptive/max | +23 pts |
| ArXivMath | 91.2% ArXivMath August 2026 release (57 problems), max effort without tools (96.9% with tools), four attempts per problem (8.9, Figures 8.9.A and B) | 82.6% Leaderboard, high | +8.6 pts |
| Terminal-Bench-Science | 58.7% Agentic scientific research Terminal-Bench-Science 0.1 | 63.3% Leaderboard, max, Claude Code | −4.6 pts |
| Harvey LAB (Legal Agent Benchmark) | 8.30% Legal Agent Benchmark (Harvey LAB) held-out set of 120 problems, all-pass rate at max effort, run by Artificial Analysis (mean criterion pass rate 91.2%) (8.14.2) | 3.75% Leaderboard | +4.6 pts |
| ProgramBench tests passed | 91.2% ProgramBench hidden test pass rate on 166 of 200 tasks, single agent, mini-swe-agent harness, no time limit (8.10.1) | 87.7% Leaderboard, xhigh, mini-SWE-agent | +3.5 pts |
| AutomationBench | 40% Business workflows AutomationBench | 42.5% Leaderboard, max, default fallbacks | −2.5 pts |
| Terminal-Bench 4.0 | 66.4% Agentic coding Terminal-Bench 4.0 (xhigh effort) | 64.8% Leaderboard, max, Claude Code | +1.6 pts |
| scBench (LatchBio SingleCellBench) | 61.2% LatchBio Bioinformatics, SingleCellBench, 195 problems, max effort (8.17.2) | 61.9% Leaderboard, published v1; LSVP, Pi | −0.7 pts |
| FrontierSWE | 62.3% FrontierSWE v2, mean of five trials per task, Proximal harness at max effort (8.7) | 62.3% Leaderboard, mean@5, proximus | Same |
| ExploitGym | 34.5% ExploitGym, 300 of 869 instances exploited with the intended vulnerability at the 6-hour budget (289 at 2 hours), mitigations disabled, counts converted to % (3.3.4, Figure 3.3.4.A) | 34.5% Leaderboard | Same |
| CursorBench 4.0 | 57.8% Agentic coding CursorBench 4.0 | 57.8% Leaderboard, max, Cursor | 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 5.5 Sep 22, 2026
Capabilities
- SWE-Bench Pro 89.9
- SWE-bench Multilingual 93.9
- SWE-bench Multimodal 61.4
- DeepSWE 74.2
- FrontierCode 1.1 54.4
- Terminal-Bench 4.0 66.4
- Terminal-Bench-Science 58.7
- FrontierSWE 62.3
- CursorBench 4.0 57.8
- ArXivMath 91.2
- ProgramBench tests passed 91.2
- Humanity's Last Exam 64.4
- DRACO 87.4
- WANDR 72.3
- Anthropic large agent teams tasks
- Chartography 64.4
- BenchCAD 73
- OSWorld 2.1 81.8
- OfficeQA 78.9
- OfficeQA Pro 67.7
- Harvey LAB (Legal Agent Benchmark) 8.3
- GDPval-AA v2.1 1846
- AA-Briefcase v1.1 1822
- Toolathlon-Verified 77.8
- AutomationBench 40
- Global MMLU 94.3
- MILU 93.1
- SpatialBench (LatchBio) 72
- scBench (LatchBio SingleCellBench) 61.2
- Morphology-to-molecule matching (Anthropic) 34
- Medicinal Chemistry (Anthropic) 63.5
- Protein Design (Anthropic)
- De novo protein binder design (Anthropic) 82.6
- Biomedical image analysis (Anthropic) 71.4
Health
- HealthBench 60.6
- HealthBench Professional 65.6
Hallucination
Safety
- Single-turn harmful request evaluations (Anthropic) 94.5
- Single-turn benign request evaluations (Anthropic) 0.03
- 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) 79.8
- Malicious computer use (Anthropic) 79.5
Robustness
- Rewind-attacker cyber adversarial robustness (Anthropic) 4
- Gray Swan IPI Arena benchmark 0.1
- Prompt injection robustness against adaptive attackers, coding (Anthropic) 54.6
- Prompt injection robustness against adaptive attackers, computer use (Anthropic) 0.07
- Prompt injection robustness against adaptive attackers, browser use (Anthropic) 0.09
Alignment
- Training distribution hack rate (Anthropic) 0.63
- Impossible tasks (Anthropic) 34.7
- Anthropic automated behavioral audit 2.49
- Anthropic undermining liberal democracy (behavioral audit metric) 1.17
- Sandbox escape evaluation (Anthropic) 1.5
- Package registry publishing evaluation (Anthropic) 33.3
- Compliance with instructions planted in pasted text (Anthropic) 7.4
- Destructive actions (Anthropic) 21
- Self-preference bias (Anthropic) 0.07
- MASK 87.4
- Silent copying of answers (Anthropic) 12.1
- Concealment of changes (Anthropic) 96.9
Monitorability
Bio and chem risk
Cyber risk
AI R&D risk
- CoBench (Anthropic) 55.8
- Anthropic ECI (AECI) 169.4