- Agentic coding SWE-bench Pro 63.2
- Agentic coding Terminal-Bench 2.1 80.4
- Multidisciplinary reasoning Humanity's Last Exam (with tools) 57.4
- Computer use OSWorld-Verified 81.2
- Knowledge work GDPval-AA v2 1618
- BrowseComp (multi agent) 86.6
- FrontierCode v1 38.8
- AutomationBench 13.5
- Best of Legal Agent Benchmark Full Public Set / Legal Agent Benchmark Harvey's Held-Out Set 8.9
- HealthBench Professional 57.8
Claude Sonnet 5
Claude Sonnet 5 ranks first on 3 of the 272 benchmarks with five or more models scored.
279
Benchmarks with a score
3
Ranked first, out of five or more models
15
In the top three
Anthropic Jun 2026
Claude Sonnet 5
Closed weights Mid-size
156.9
#33 of 256
Epoch Capabilities Index
Self-Reported Index 156.3 #53 of 404
- DeepSWE 53.9% #18 /28
- AutomationBench 10.7% #28 /42
- Terminal-Bench 4.0 12.4% #32 /41
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 | |
|---|---|---|---|---|
| Phare | 78.1% | #3 of 62 | Official leaderboard | |
| BullshitBench V2 | 80.8% | #9 of 129 | Official leaderboard | |
| Humanity's Last Exam | 57.4% | #18 of 154 | Lab model card | |
| INCLUDE | 86.5% | #4 of 31 | Lab model card | |
| BullshitBench | 65.5% | #14 of 107 | Official leaderboard | |
| LMArena (Elo) | 1,461 | #50 of 335 | Official leaderboard | |
| Longform Creative Writing (EQ-Bench) | 78.3 | #20 of 119 | Official leaderboard | |
| Continual Learning Exam | 40.2% | #2 of 11 | Official leaderboard | |
| Creative Writing v3 | 76.3 | #28 of 120 | Official leaderboard | |
| Structured Output Benchmark | 86.2% | #5 of 19 | Official leaderboard | |
| SEA-HELM | 73.8 | #21 of 71 | Official leaderboard | |
| OpenAI MRCR v2 | 52.2% | #17 of 47 | Official leaderboard | |
| Judgemark v4 | 70.9 | #15 of 40 | Official leaderboard | |
| StudyArena Text | 1,242 | #32 of 85 | Official leaderboard | |
| DesignArena ASCII Art | 1,200 | #30 of 77 | Official leaderboard | |
| Document Arena | 1,469 | #16 of 39 | Official leaderboard | |
| EQ-Bench 4 | 1,221 | #14 of 32 | Official leaderboard | |
| Extended NYT Connections | 75.1% | #51 of 103 | Official leaderboard | |
| SimpleQA Verified | 32.1% | #28 of 56 | Official leaderboard | |
| AA-LCR Extended | 61.8% | #27 of 49 | Official leaderboard | |
| Global MMLU | 89% | #8 of 14 | Lab model card | |
| HLE Extended | 29.1% | #18 of 31 | Official leaderboard | |
| Corp Speak Arena | 1,462 | #14 of 22 | Official leaderboard | |
| ELI5 Arena | 1,416 | #14 of 18 | Official leaderboard | |
| MILU | 89.3% | #9 of 11 | Lab model card | |
| FACTS Parametric V2 | 20.5% | #24 of 27 | Official leaderboard | |
| Joke Arena | 1,418 | #24 of 25 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ProofBench | 66% | #7 of 59 | Official leaderboard | |
| GPQA Diamond | 90.5% | #45 of 229 | Epoch AI run | |
| DTBench | 92.5% | #36 of 180 | Official leaderboard | |
| OTIS Mock AIME 2024-2025 | 94.7% | #51 of 212 | Epoch AI run | |
| ProofBench v1.1 | 77% | #17 of 52 | Official leaderboard | |
| SimpleBench | 60.6% | #41 of 106 | Official leaderboard | |
| Surface Evolver Bench | 60% | #10 of 25 | Official leaderboard | |
| Organic Chemistry (Anthropic) | 81% | #2 of 5 | Lab model card | |
| FrontierMath Tiers 1-3 v2 | 65.6% | #36 of 86 | Official leaderboard | |
| FrontierMath Tier 4 v2 | 29.3% | #35 of 68 | Official leaderboard | |
| ForecastBench | 0.14 | #47 of 82 | Official leaderboard | |
| scBench (LatchBio SingleCellBench) | 55.6% | #15 of 26 | Official leaderboard | |
| LatchBio Capabilities v1.0 | 40.9% | #16 of 27 | Official leaderboard | |
| Verifying the Verifiers (Project APE CRED benchmark) | 77.5% | #15 of 23 | Official leaderboard | |
| USAMO 2026 | 79.5% | #4 of 6 | Lab model card | |
| SpatialBench (LatchBio) | 70% | #5 of 7 | Lab model card | |
| BTF-3 | 0.14 | #6 of 8 | Official leaderboard | |
| ArXivMath | 65.7% | #8 of 9 | Lab model card | |
| Structural Biology (Anthropic) | 69% | #2 of 2 | Lab model card |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SWE-bench Verified | 85.2% | #6 of 155 | Lab model card | |
| SWE Atlas Codebase QnA Extended | 32.7% | #7 of 38 | Official leaderboard | |
| WeirdML | 69% | #22 of 112 | Official leaderboard | |
| DesignArena Game Dev | 1,271 | #35 of 144 | Official leaderboard | |
| DesignArena Website | 1,268 | #37 of 152 | Official leaderboard | |
| DesignArena Code | 1,267 | #36 of 144 | Official leaderboard | |
| SWE-bench Verified Extended | 67.3% | #11 of 44 | Official leaderboard | |
| MirrorCode | 63.3% | #10 of 39 | Epoch AI run | |
| Vibe Code Bench | 81.3% | #31 of 120 | Official leaderboard | |
| SWE-Bench Pro | 63.2% | #21 of 81 | Lab model card | |
| SWE-bench Multilingual | 78.3% | #19 of 73 | Lab model card | |
| DesignArena Android (Kotlin) | 1,211 | #12 of 46 | Official leaderboard | |
| DesignArena Full Stack | 1,222 | #14 of 50 | Official leaderboard | |
| AndroidBench | 76.2% | #9 of 32 | Official leaderboard | |
| ALE-Bench | 1,783 | #37 of 126 | Official leaderboard | |
| Code Migration | 44.4% | #23 of 78 | Official leaderboard | |
| DesignArena UI Component | 1,267 | #42 of 140 | Official leaderboard | |
| GSO-Bench | 37.3% | #9 of 30 | Official leaderboard | |
| DesignArena 3D Design | 1,237 | #42 of 137 | Official leaderboard | |
| WebDev Arena | 1,537 | #34 of 105 | arena.ai, via Epoch AI | |
| GBAEval | 65.3% | #8 of 24 | Official leaderboard | |
| SWE-Bench Pro V2 Hard | 88.2% | #6 of 18 | Official leaderboard | |
| DesignArena Frontend Web App | 1,217 | #18 of 53 | Official leaderboard | |
| DesignArena Agentic Overall | 1,216 | #26 of 76 | Official leaderboard | |
| DesignArena Data Visualization | 1,247 | #42 of 119 | Official leaderboard | |
| APEX-SWE | 46.4% | #24 of 65 | Official leaderboard | |
| DesignArena Agentic Game Dev | 1,166 | #11 of 29 | Official leaderboard | |
| Code Arena | 1,539 | #49 of 128 | Official leaderboard | |
| DesignArena React Native | 1,202 | #22 of 57 | Official leaderboard | |
| Senior SWE-Bench | 28% | #9 of 23 | Official leaderboard | |
| DesignArena SVG | 1,184 | #47 of 105 | Official leaderboard | |
| SWE-rebench | 56.8% | #6 of 13 | Official leaderboard | |
| FrontierCode 1.1 | 42.7% | #4 of 8 | Official leaderboard | |
| Quadruped reinforcement learning task (Anthropic) | 19.9 | #4 of 8 | Lab model card | |
| SWE Refactor Bench | 15% | #4 of 8 | Official leaderboard | |
| DesignArena Godot Game Dev | 1,198 | #13 of 24 | Official leaderboard | |
| Image-to-WebDev Arena | 1,539 | #31 of 55 | Official leaderboard | |
| ReactBench | 32.3% | #16 of 28 | Official leaderboard | |
| DesignArena Image to Frontend | 1,182 | #18 of 31 | Official leaderboard | |
| DeepSWE | 53.9% | #18 of 28 | Official leaderboard | |
| Landing Page Arena | 1,423 | #10 of 14 | Official leaderboard | |
| DesignArena Image to HTML | 1,172 | #37 of 51 | Official leaderboard | |
| DAXBench v2 | 70.4% | #35 of 48 | Official leaderboard | |
| IOI v2 (Vals) | 45% | #33 of 45 | Official leaderboard | |
| FrontierCode | 38.8% | #3 of 4 | Lab model card | |
| Novel compiler task (Anthropic) | 76.1% | #3 of 4 | Lab model card | |
| Time series forecasting task (Anthropic) | 5.80 | #7 of 9 | Lab model card | |
| SWE-Bench Pro V2 | 93.2% | #15 of 18 | Official leaderboard | |
| Vibe Code Bench 1-100 | 13.8% | #22 of 26 | Official leaderboard | |
| MLS-Bench | 31.4% | #13 of 15 | Official leaderboard | |
| DesignArena Blender | 990 | #13 of 14 | Official leaderboard | |
| Next.js Agent Evals | 82% | #13 of 14 | Official leaderboard | |
| CursorBench | 34.1% | #20 of 21 | cursor.com, via Epoch AI | |
| CursorBench 4.0 | 34.1% | #20 of 21 | Official leaderboard | |
| SWE-bench Multimodal | 28.1% | #12 of 12 | Lab model card | |
| CursorBench 3.1 | 61.2% | #5 of 5 | Lab model card | |
| τ^τ-bench | 14.9% | #5 of 5 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LisanBench | 5724.0 | #21 of 133 | Official leaderboard | |
| OSWorld-Verified | 81.2% | #8 of 41 | Lab model card | |
| Long-Horizon-Terminal-Bench (LHTB) | 49.7% | #5 of 25 | Official leaderboard | |
| Game Arena Five in a Row | 865 | #4 of 20 | Official leaderboard | |
| BrowseComp | 86.6% | #17 of 81 | Lab model card | |
| Toolathlon | 54.3% | #7 of 29 | Lab model card | |
| LLM Chess | 806 | #31 of 123 | Official leaderboard | |
| Mystery Game Puzzles | 35% | #25 of 84 | Epoch AI run | |
| Toolathlon-Verified | 71.6% | #8 of 25 | Official leaderboard | |
| Chess Puzzles | 45% | #31 of 93 | Epoch AI run | |
| Game Arena Go | 1,118 | #13 of 39 | Official leaderboard | |
| DuetBench | 1,022 | #3 of 9 | Official leaderboard | |
| Era by Eon | 67.7% | #3 of 9 | Official leaderboard | |
| Game Arena Nine Men's Morris | 750 | #15 of 39 | Official leaderboard | |
| Agent Arena | 2.00% | #23 of 56 | Official leaderboard | |
| Game Arena Dots and Boxes | 931 | #17 of 40 | Official leaderboard | |
| Game Arena Checkers | 954 | #19 of 41 | Official leaderboard | |
| RuneBench | 4.81 | #36 of 76 | Official leaderboard | |
| CEO-Bench | $64,459 | #9 of 19 | Official leaderboard | |
| CLASHarness | 20.8% | #7 of 14 | Official leaderboard | |
| ApprenticeBench | 16% | #10 of 18 | Official leaderboard | |
| Game Arena Clobber | 247 | #23 of 39 | Official leaderboard | |
| Game Arena Four in a Row | 322 | #26 of 43 | Official leaderboard | |
| OSWorld-Science | 26.3% | #7 of 11 | Official leaderboard | |
| GameHorizon | 64.7 | #24 of 37 | Official leaderboard | |
| LMArena Search Arena | 1,191 | #24 of 37 | Official leaderboard | |
| Game Arena Dark Hex | 206 | #27 of 41 | Official leaderboard | |
| Terminal-Bench 2.1 Extended | 28.9% | #29 of 44 | Official leaderboard | |
| AutomationBench | 10.7% | #28 of 42 | Official leaderboard | |
| BrowseComp Extended | 19% | #28 of 42 | Official leaderboard | |
| Kaggle Game Arena | 260 | #28 of 42 | Official leaderboard | |
| BabaIsBench | 67% | #22 of 32 | Official leaderboard | |
| Game Arena Ultimate Tic-Tac-Toe | 186 | #27 of 39 | Official leaderboard | |
| Terminal-Bench 2.1 | 74.6% | #9 of 13 | Official leaderboard | |
| Game Arena Reversi | 253 | #28 of 39 | Official leaderboard | |
| Game Arena Poker | -7.27 | #23 of 31 | Official leaderboard | |
| Terminal-Bench 4.0 | 12.4% | #32 of 41 | Official leaderboard | |
| Game Arena Chess | 587 | #34 of 42 | Official leaderboard | |
| Learn2Play Bench | 55.9 | #9 of 11 | Official leaderboard | |
| Terminal-Bench 3.0 | 14.6% | #9 of 11 | Official leaderboard | |
| Game Arena Coin Game | 237 | #32 of 39 | Official leaderboard | |
| Game Arena Lines of Action | 222 | #32 of 39 | Official leaderboard | |
| Terminal-Bench 4.1 | 12.1% | #39 of 46 | Official leaderboard | |
| Game Arena Word Art | 146 | #34 of 39 | Official leaderboard | |
| CAR-bench | 43.3% | #11 of 12 | Official leaderboard | |
| Game Arena Bargaining | 82 | #37 of 39 | Official leaderboard | |
| Game Arena Word Association | 25 | #38 of 40 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ProfBench | 60.4% | #3 of 72 | Official leaderboard | |
| MortgageTax | 70% | #4 of 87 | Official leaderboard | |
| TaxEval v2 | 75.6% | #15 of 129 | Official leaderboard | |
| CorpFin v2 | 67% | #15 of 119 | Official leaderboard | |
| RCM-Bench-1 | 91.2% | #4 of 22 | Official leaderboard | |
| Vals Index v2 | 59.6% | #14 of 65 | Official leaderboard | |
| LMBuild | 62.5% | #5 of 18 | Official leaderboard | |
| Excel Modeling Benchmark | 66.3% | #22 of 76 | Official leaderboard | |
| Legal Research Bench | 41.8% | #30 of 93 | Official leaderboard | |
| LegalBench | 83.9% | #46 of 138 | Official leaderboard | |
| MedCode | 47.5% | #32 of 96 | Official leaderboard | |
| Vending-Bench 2 | $6,378 | #25 of 69 | Official leaderboard | |
| DesignArena PDF | 1,211 | #14 of 38 | Official leaderboard | |
| Vals Finance Agent v2 | 53.9% | #30 of 80 | Official leaderboard | |
| PatientAgentBench | 4.20 | #3 of 8 | Official leaderboard | |
| DesignArena HTML Slides | 1,112 | #11 of 28 | Official leaderboard | |
| Public Benefits Bench | 66% | #20 of 50 | Official leaderboard | |
| GDPval-AA v2 | 1,618 | #11 of 27 | Lab model card | |
| EnterpriseBench: CoreCraft | 51.3% | #21 of 51 | Official leaderboard | |
| CHI-Bench | 20% | #10 of 23 | Official leaderboard | |
| Tax Agent Bench | 29.4% | #37 of 84 | Official leaderboard | |
| APEX-Agents | 54.5% | #27 of 61 | Collected by Epoch AI, source not named | |
| HealthBench | 58.7% | #14 of 31 | Lab model card | |
| DesignArena Python-PPTX Slides | 1,180 | #17 of 36 | Official leaderboard | |
| YC-Bench | $1,163,518 | #28 of 59 | Official leaderboard | |
| Harvey LAB (Legal Agent Benchmark) | 5.00% | #38 of 80 | Official leaderboard | |
| GDPval Extended | 1.10% | #21 of 43 | Official leaderboard | |
| Vals Index v2.1 | 51.9% | #24 of 49 | Official leaderboard | |
| APEX-Agents 1.1 | 67.4% | #32 of 65 | Official leaderboard | |
| LifeSci Diamond | 1.80% | #10 of 18 | Official leaderboard | |
| DesignArena Spreadsheet | 1,046 | #22 of 39 | Official leaderboard | |
| Tuesday Work Index | 46.1 | #25 of 44 | Official leaderboard | |
| DAYJOB: Healthcare | 2.80% | #22 of 37 | Official leaderboard | |
| OfficeQA | 73.3% | #9 of 14 | Lab model card | |
| TRLawBench | 8.92 | #22 of 34 | Official leaderboard | |
| DAYJOB: Finance | 4.50% | #26 of 39 | Official leaderboard | |
| Holdings-IndividualTaxBench | 71.5% | #9 of 13 | Official leaderboard | |
| APEX-Accounting | 55.7% | #34 of 49 | Official leaderboard | |
| GDP.xlsx | 16.9% | #14 of 20 | Official leaderboard | |
| AA-Briefcase | 1,393 | #5 of 7 | Lab model card | |
| HealthBench Professional | 57.8% | #19 of 26 | Lab model card | |
| MedScribe | 76% | #71 of 97 | Official leaderboard | |
| RealmPathology | 71.7% | #10 of 13 | Official leaderboard | |
| EEBench | 40.3% | #24 of 31 | Official leaderboard | |
| TaxCalcBench v2 | 6.00% | #28 of 34 | Official leaderboard | |
| OfficeQA Pro | 59.4% | #14 of 16 | Lab model card | |
| MedXpertQA MM Extended | 43.8% | #38 of 41 | Official leaderboard | |
| APEX-1 | 49.6% | #68 of 71 | Official leaderboard | |
| PolicyBench | 72.3% | #49 of 51 | Official leaderboard | |
| BenchCAD | 26.6% | #11 of 11 | Lab model card | |
| Real-World Finance v2 (Anthropic) | 1,219 | #3 of 3 | Lab model card |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GeoGuesser environment (FineEnvs) | 0.64 | #1 of 14 | Official leaderboard | |
| GDP.pdf | 67.5% | #3 of 17 | Lab model card | |
| Vision Arena | 1,261 | #38 of 136 | Official leaderboard | |
| MineBench | 1,611 | #30 of 75 | Official leaderboard | |
| ChartMuseum | 70.1% | #2 of 5 | Lab model card | |
| vCX Benchmark | 87.5% | #18 of 42 | Official leaderboard | |
| ZeroBench | 13% | #33 of 76 | Official leaderboard | |
| GeoBench | 74% | #22 of 48 | Official leaderboard | |
| CharXiv | 77% | #46 of 75 | Lab model card | |
| BVB | 64.6 | #29 of 47 | Official leaderboard | |
| ParseBench | 61.2% | #28 of 45 | Official leaderboard | |
| Chartography | 16.6% | #28 of 43 | Official leaderboard | |
| VoxelBench | 1,546 | #37 of 56 | Official leaderboard | |
| ArtifactArena | 986 | #14 of 21 | Official leaderboard | |
| Blueprint-Bench 2 | 24.9% | #28 of 38 | Official leaderboard | |
| DesignArena Code to Video | 1,102 | #23 of 31 | Official leaderboard | |
| Humanity's Sixth Sense | 24.8% | #28 of 34 | Official leaderboard | |
| RoTA-Bench | 0% | #7 of 8 | Official leaderboard | |
| CharXiv Extended | 71.1% | #34 of 38 | Official leaderboard | |
| MMMU-Pro Extended | 20.7% | #36 of 38 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BioTIER Refuse | 96.1% | #4 of 69 | Official leaderboard | |
| Molecular Biology Capabilities Test | 59.2% | #9 of 90 | Official leaderboard | |
| Human Pathogen Capabilities Test | 66.5% | #12 of 86 | Official leaderboard | |
| MASK | 96.9% | #3 of 19 | Lab model card | |
| BrowseSafe-Adapt | 30.7% | #1 of 6 | Official leaderboard | |
| LinuxArena | 0% | #1 of 6 | Lab model card | |
| Malicious use of Claude Code (Anthropic) | 92.4% | #3 of 17 | Lab model card | |
| nocot-bench (NCRI) | 90.8 | #36 of 185 | Official leaderboard | |
| Human Pathogen Capabilities Test v2 (HPCT v2) | 65% | #14 of 71 | Official leaderboard | |
| BrowseSafe++ | 88.4% | #5 of 15 | Official leaderboard | |
| Protocols (Anthropic life sciences) | 62% | #1 of 3 | Lab model card | |
| MetagenomicsBench | 42.3% | #9 of 26 | Official leaderboard | |
| Gray Swan Agent Red Teaming (ART) benchmark | 0.10% | #5 of 14 | Lab model card | |
| Dictatorship Eval | 83.5% | #8 of 20 | Official leaderboard | |
| VariantBench | 39.8% | #11 of 27 | Official leaderboard | |
| Prompt injection robustness against adaptive attackers, coding (Anthropic) | 0.31% | #6 of 14 | Lab model card | |
| World-Class Bio | 49% | #32 of 74 | Official leaderboard | |
| SHADE-Arena | 10% | #7 of 16 | Lab model card | |
| DrugDiscoveryBench | 50% | #10 of 21 | Official leaderboard | |
| TxBench-Oligo-Discovery | 36.6% | #16 of 30 | Official leaderboard | |
| BioSecBench-Function | 33.8% | #16 of 27 | Official leaderboard | |
| scBench-Long | 15.9% | #16 of 27 | Official leaderboard | |
| Prompt injection robustness against adaptive attackers, browser use (Anthropic) | 0.93% | #8 of 13 | Lab model card | |
| EpiBench | 21.7% | #17 of 27 | Official leaderboard | |
| TxBench-Preclinical-Pharmacology | 49.3% | #19 of 30 | Official leaderboard | |
| Gray Swan IPI Arena benchmark | 0.60% | #7 of 11 | Lab model card | |
| OncologyBench | 14.7% | #20 of 30 | Official leaderboard | |
| Anthropic automated behavioral audit | 2.53 | #4 of 6 | Lab model card | |
| Prompt injection live bug bounty (Anthropic with Gray Swan) | 0.19% | #2 of 3 | Lab model card | |
| SpatialBench-Long | 24.4% | #19 of 27 | Official leaderboard | |
| Anthropic political even-handedness eval | 96.4% | #10 of 14 | Lab model card | |
| Malicious computer use (Anthropic) | 84.7% | #10 of 14 | Lab model card | |
| DistressBench | 68% | #31 of 43 | Official leaderboard | |
| DeepsecBench | 14.9 | #21 of 29 | Official leaderboard | |
| CyberBench v1.1 (Vals) | 61.9% | #32 of 44 | Official leaderboard | |
| Single-turn harmful request evaluations (Anthropic) | 96.7% | #14 of 19 | Lab model card | |
| BioSecBench-Surveillance | 39% | #21 of 27 | Official leaderboard | |
| Prompt injection robustness against adaptive attackers, computer use (Anthropic) | 2.25% | #11 of 14 | Lab model card | |
| TxBench-Antibody-Discovery | 32% | #24 of 30 | Official leaderboard | |
| BBQ (Bias Benchmark for QA) | 72.4% | #20 of 25 | Lab model card | |
| DNA Synthesis Screening Evasion | 50% | #4 of 5 | Lab model card | |
| Firefox 147 exploit evaluation (Anthropic) | 13.2% | #4 of 5 | Lab model card | |
| Virology Capabilities Test (VCT) | 45% | #25 of 31 | Lab model card | |
| Anthropic undermining liberal democracy (behavioral audit metric) | 1.22 | #5 of 6 | Lab model card | |
| Binary Exploitation Benchmark (Anthropic, OSS-Fuzz based) | 54.5% | #5 of 6 | Lab model card | |
| SecIT Bench | 65.9% | #22 of 26 | Official leaderboard | |
| CyberGym | 52.7% | #29 of 34 | Lab model card | |
| SRE-Bench | 0.38% | #30 of 35 | Official leaderboard | |
| Single-turn benign request evaluations (Anthropic) | 0.59% | #18 of 19 | Lab model card | |
| ExploitBench | 31% | #20 of 21 | Lab model card | |
| BioTIER Permit | 78.3% | #66 of 69 | Official leaderboard | |
| Uncritically reporting flawed results (Anthropic) | 8.60 | #3 of 3 | Lab model card |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Political Science LLM Benchmark | 69.7% | #3 of 42 | Official leaderboard | |
| ACCoRD | 76.9 | #25 of 152 | Official leaderboard | |
| LMCA | 50% | #29 of 147 | Official leaderboard | |
| Conceptual Reasoning Index | 61.8 | #33 of 151 | Official leaderboard | |
| Vals Multimodal Index | 68.8% | #8 of 33 | Official leaderboard | |
| SAGE (Vals) | 48.9% | #24 of 81 | Official leaderboard |
Its lab's numbers put it at 156.3, in line with its ECI
The Self-Reported Index fits one score to the 22 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 152.5 to 160.1. How we compute it.
- MILU 89.3
- Global MMLU 89
- BrowseComp 86.6
- INCLUDE 86.5
- SWE-bench Verified 85.2
- OSWorld-Verified 81.2
- Terminal-Bench 2.1 80.4
- SWE-bench Multilingual 78.3
- CharXiv 77
- OfficeQA 73.3
- GDP.pdf 67.5
- ArXivMath 65.7
- SWE-Bench Pro 63.2
- OfficeQA Pro 59.4
- HealthBench 58.7
- HealthBench Professional 57.8
- Humanity's Last Exam 57.4
- Toolathlon 54.3
- SWE-bench Multimodal 28.1
- BenchCAD 26.6
- AutomationBench 13.5
- Harvey LAB (Legal Agent Benchmark) 8.9
Its lab's numbers next to 4 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 |
|---|---|---|---|
| Terminal-Bench 2.1 | 80.4% Agentic coding Terminal-Bench 2.1 | 74.6% Leaderboard, high, Claude Code | +5.8 pts |
| Harvey LAB (Legal Agent Benchmark) | 8.90% Best of Legal Agent Benchmark Full Public Set / Legal Agent Benchmark Harvey's Held-Out Set | 5.00% Leaderboard | +3.9 pts |
| AutomationBench | 13.5% AutomationBench | 10.7% Leaderboard, max | +2.8 pts |
| scBench (LatchBio SingleCellBench) | 56% LatchBio SingleCellBench, 195 problems, score 0.56 (8.14.2, Figure 8.14.6.A) | 55.6% Leaderboard, published v1, Pi | +0.4 pts |
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. 67 of them aren't in the launch post.
System Card: Claude Sonnet 5 Jun 30, 2026
Capabilities
- SWE-bench Verified 85.2
- SWE-Bench Pro 63.2
- SWE-bench Multilingual 78.3
- SWE-bench Multimodal 28.1
- Terminal-Bench 2.1 80.4
- FrontierCode 38.8
- CursorBench 3.1 61.2
- USAMO 2026 79.5
- ArXivMath 65.7
- ProgramBench tests passed
- Humanity's Last Exam 43.2
- BrowseComp 84.7
- GDP.pdf 67.5
- OSWorld-Verified 81.2
- BenchCAD 26.6
- ChartMuseum 70.1
- CharXiv 77
- OfficeQA 73.3
- OfficeQA Pro 59.4
- Real-World Finance v2 (Anthropic) 1219
- Harvey LAB (Legal Agent Benchmark) 5.8
- GDPval-AA v2 1618
- Toolathlon 54.3
- AutomationBench 13.5
- AA-Briefcase 1393
- Global MMLU 89
- MILU 89.3
- INCLUDE 86.5
- BioMysteryBench
- SpatialBench (LatchBio) 70
- scBench (LatchBio SingleCellBench) 56
- Structural Biology (Anthropic) 69
- ProteinGym 37
- Organic Chemistry (Anthropic) 81
- Protocols (Anthropic life sciences) 62
Health
- HealthBench 58.7
- HealthBench Professional 57.8
Hallucination
Safety
- Single-turn harmful request evaluations (Anthropic) 96.7
- Single-turn benign request evaluations (Anthropic) 0.59
- 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) 92.4
- Malicious computer use (Anthropic) 84.7
Robustness
- Gray Swan Agent Red Teaming (ART) benchmark 0.1
- Gray Swan IPI Arena benchmark 0.6
- Prompt injection live bug bounty (Anthropic with Gray Swan) 0.19
- Prompt injection robustness against adaptive attackers, coding (Anthropic) 0.31
- Prompt injection robustness against adaptive attackers, computer use (Anthropic) 2.25
- Prompt injection robustness against adaptive attackers, browser use (Anthropic) 0.93