Anthropic · model
Claude Haiku 5.5 benchmark scores
As of 2026-11-22, Claude Haiku 5.5 (Anthropic) has measured scores on 49 benchmarks and estimated scores on 113 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 90.9% | estimated ± 2.7 pp, high confidence |
| AA-HLE | 44.4% | measured |
| AA-SciCode | 55.0% | measured |
| AIME26 | 99.2% | estimated ± 1.2 pp, low confidence |
| Artificial Analysis Intelligence Index | 36.5% | measured |
| CritPt | 18.9% | measured |
| GDPval-AA | 56.0% | measured |
| GPQA Diamond | 90.6% | estimated ± 1.4 pp, high confidence |
| HLE | 54.3% | estimated ± 2.2 pp, high confidence |
| AA-LCR | 82.7% | measured |
| LiveCodeBench v6 | 95.5% | estimated ± 4.9 pp, medium confidence |
| AA-Omniscience Accuracy | 34.7% | measured |
| SciCode | 54.7% | estimated ± 3.1 pp, high confidence |
| Terminal-Bench 2.1 | 89.1% | estimated ± 5.4 pp, low confidence |
| BrowseComp | 82.2% | estimated ± 7.1 pp, medium confidence |
| HLE w/ tools | 57.4% | measured |
| LongBench v2 | 68.1% | estimated ± 5.5 pp, medium confidence |
| τ²-bench results | 98.1% | estimated ± 3.1 pp, high confidence |
| AA Coding Index | 79.7% | estimated ± 1.2 pp, high confidence |
| AA-MMMU-Pro | 79.6% | estimated ± 5.4 pp, medium confidence |
| APEX-Agents | 50.4% | estimated ± 9.4 pp, low confidence |
| APEX-Agents-AA | 37.7% | estimated ± 9.3 pp, medium confidence |
| BioMysteryBench (human-difficult) | 47.8% | estimated ± 6.8 pp, low confidence |
| DeepSearchQA | 89.6% | estimated ± 9.6 pp, medium confidence |
| SWE-bench Verified | 90.4% | estimated ± 2.2 pp, medium confidence |
| BFCL v4 | 77.1% | estimated ± 2.7 pp, medium confidence |
| CyberGym | 88.2% | estimated ± 6.0 pp, medium confidence |
| JobBench | 60.8% | estimated ± 8.7 pp, medium confidence |
| SWE-bench Pro | 75.7% | estimated ± 4.2 pp, high confidence |
| WideResearch | 83.0% | estimated ± 3.3 pp, medium confidence |
| AA Agentic Index | 52.1% | estimated ± 3.1 pp, high confidence |
| MMLU-Pro | 85.8% | estimated ± 2.5 pp, low confidence |
| FailureSensorIQ | 66.3% | estimated ± 1.7 pp, high confidence |
| GPQA | 93.1% | estimated ± 0.9 pp, high confidence |
| MMLU | 90.5% | estimated ± 0.9 pp, medium confidence |
| Vals GPQA Diamond | 93.0% | estimated ± 1.4 pp, high confidence |
| Vals LiveCodeBench | 87.1% | estimated ± 1.9 pp, high confidence |
| Vals MMLU-Pro | 89.0% | estimated ± 1.6 pp, medium confidence |
| MMMLU | 90.7% | estimated ± 0.7 pp, low confidence |
| Chess Puzzles | 55.0% | measured |
| AA MMLU-Pro | 86.3% | estimated ± 0.8 pp, high confidence |
| Gert Labs | 58.6% | measured |
| AA-AnalystAgent | 53.8% | measured |
| AA Harvey LAB v1.0 | 89.9% | measured |
| AA Tau3 Banking | 45.5% | estimated ± 4.1 pp, high confidence |
| ApprenticeBench | 13.8% | estimated ± 10.4 pp, low confidence |
| ARC-AGI-1 | 89.0% | measured |
| ARC-AGI-2 | 61.3% | measured |
| BabaIsBench | 99.3% | estimated ± 1.2 pp, high confidence |
| CursorBench 3.1 | 56.3% | estimated ± 3.4 pp, medium confidence |
| CursorBench 3.2 | 63.4% | estimated ± 1.6 pp, medium confidence |
| EBR-bench | 43.8% | measured |
| Terminal-Bench 3.0 | 33.4% | estimated ± 3.2 pp, high confidence |
| FrontierCode 1.1 Main | 46.4% | measured |
| FrontierMath v2 (Tier 4) | 46.3% | measured |
| FrontierMath v2 (Tiers 1-3) | 75.1% | measured |
| FrontierSWE v2 | 43.8% | measured |
| Furniture Assembly | 47.5% | measured |
| MaxxBench | 65.7% | estimated ± 4.3 pp, medium confidence |
| MazeBench | 1.0% | measured |
| MirrorCode | 78.9% | measured |
| Mystery Game Puzzles | 30.0% | measured |
| NYT Connections | 65.7% | measured |
| OfficeQA Pro | 51.9% | estimated ± 7.0 pp, low confidence |
| OSWorld-Verified | 85.1% | estimated ± 8.4 pp, low confidence |
| ParseBench-1 | 56.1% | measured |
| Terminal-Bench 4.1 | 40.0% | measured |
| Terminal-Bench Hard | 63.6% | estimated ± 2.8 pp, medium confidence |
| Vals SWE-bench | 86.6% | estimated ± 2.8 pp, high confidence |
| Terminal-Bench 2.1 (Vals AI) | 72.8% | estimated ± 6.9 pp, low confidence |
| VulcanBench v3 | 78.2% | estimated ± 3.9 pp, medium confidence |
| CWE-Bench-AA | 63.3% | estimated ± 3.8 pp, high confidence |
| AA Cyber Index | 41.4% | estimated ± 12.4 pp, low confidence |
| DeepsecBench-AA | 0.0% | estimated ± 7.5 pp, low confidence |
| Harvey LAB-AA v1.1 | 2.3% | estimated ± 3.5 pp, low confidence |
| HealthBench Professional (length-adjusted, Artificial Analysis) | 58.4% | measured |
| HealthBench Professional (raw, Artificial Analysis) | 64.8% | measured |
| AA ITBench | 46.6% | estimated ± 4.6 pp, medium confidence |
| Bug Hunt Bench | 21.5% | measured |
| Chartography (no tools) | 46.4% | measured |
| Chartography (tools) | 86.3% | estimated ± 2.7 pp, medium confidence |
| CursorBench 4.0 | 48.4% | measured |
| CWE-bench v1 | 57.4% | estimated ± 5.1 pp, medium confidence |
| DeepSWE | 72.2% | estimated ± 4.1 pp, medium confidence |
| DRACO | 86.9% | estimated ± 0.7 pp, low confidence |
| FrontierCode 2.0 | 28.9% | measured |
| FrontierPhysics | 4.2% | measured |
| HLE w/o tools | 45.9% | measured |
| OSWorld 2.1 | 72.3% | estimated ± 3.6 pp, high confidence |
| OSWorld 2.1 Offline | 72.4% | measured |
| PostTrainBench v1.1 | 35.7% | estimated ± 1.7 pp, medium confidence |
| SpireClimb | 5.8% | estimated ± 7.2 pp, low confidence |
| StealthMark | 49.5% | measured |
| SWE-bench Multimodal | 2.3% | estimated ± 5.6 pp, low confidence |
| Terminal-Bench 4.0 | 39.2% | measured |
| Terminal-Bench-Science 0.1 | 25.7% | estimated ± 10.7 pp, low confidence |
| Toolathlon-Verified | 77.7% | estimated ± 1.8 pp, medium confidence |
| VulcanBench Frontier v4 | 92.2% | measured |
| AA AutomationBench | 35.4% | measured |
| GDP.pdf | 20.8% | measured |
| AA Terminal-Bench 4.0 | 32.8% | measured |
| ARC-AGI-3 | 0.5% | measured |
| ARC-AGI-3 Provider Adapter | 2.8% | measured |
| cursorBench41 | 48.4% | measured |
| CWE-bench | 46.1% | estimated ± 4.3 pp, medium confidence |
| MKA | 41.0% | measured |
| SkillsBench official leaderboard (BenchFlow) | 55.3% | estimated ± 6.1 pp, medium confidence |
| BinaryAudit | 35.8% | estimated ± 12.1 pp, low confidence |
| CompileBench | 100.0% | estimated ± 10.0 pp, low confidence |
| OTelBench | 32.1% | estimated ± 4.1 pp, low confidence |
| Vibe Code Bench | 75.2% | estimated ± 5.6 pp, low confidence |
| FACTS Multimodal | 38.9% | measured |
| Hermes Bench | 71.0% | measured |
| Hermes Index | 37.0% | measured |
| ParseBench-2 | 71.2% | measured |
| CharXiv | 82.3% | estimated ± 5.6 pp, low confidence |
| CharXiv w/o tools | 75.1% | estimated ± 1.8 pp, low confidence |
| ExploitBench | 33.2% | estimated ± 6.5 pp, medium confidence |
| ExploitGym | 15.8% | estimated ± 2.7 pp, high confidence |
| C-Eval | 96.5% | estimated ± 1.0 pp, low confidence |
| HMMT Feb 2026 | 100.0% | estimated ± 1.5 pp, low confidence |
| HMMT Nov 2025 | 95.7% | estimated ± 1.5 pp, low confidence |
| MCP Atlas | 82.2% | estimated ± 10.7 pp, low confidence |
| MMAnswerBench | 93.0% | estimated ± 2.4 pp, low confidence |
| MMLU-Redux | 95.3% | estimated ± 0.8 pp, medium confidence |
| MMMU-Pro | 78.9% | estimated ± 5.3 pp, low confidence |
| NL2Repo | 56.1% | estimated ± 6.5 pp, low confidence |
| ScreenSpot Pro | 78.6% | estimated ± 10.1 pp, low confidence |
| SuperGPQA | 73.1% | estimated ± 9.3 pp, medium confidence |
| Toolathlon | 55.9% | estimated ± 6.3 pp, low confidence |
| HealthBench Hard | 32.3% | estimated ± 8.6 pp, low confidence |
| MedXpertQA (Text) | 63.2% | estimated ± 8.9 pp, low confidence |
| React Native Evals | 78.8% | estimated ± 4.7 pp, medium confidence |
| ResearchClawBench | 21.8% | estimated ± 2.2 pp, low confidence |
| PostTrainBench | 33.1% | estimated ± 2.4 pp, high confidence |
| DAYJOB | 1.4% | estimated ± 2.7 pp, high confidence |
| DAYJOB Healthcare | 3.1% | estimated ± 4.0 pp, high confidence |
| Finance Agent v2 | 56.9% | estimated ± 2.4 pp, medium confidence |
| BioMysteryBench (human-solvable) | 88.9% | estimated ± 0.6 pp, medium confidence |
| FrontierCode 1.1 Extended | 61.4% | estimated ± 3.0 pp, medium confidence |
| HealthBench (raw) | 52.7% | estimated ± 6.0 pp, medium confidence |
| HealthBench (length-adjusted) | 53.8% | estimated ± 0.9 pp, medium confidence |
| HealthBench Professional | 59.4% | estimated ± 2.1 pp, medium confidence |
| HealthBench Professional (raw) | 59.9% | estimated ± 2.3 pp, medium confidence |
| HLE-Verified | 54.0% | estimated ± 6.8 pp, low confidence |
| LABBench2 | 81.7% | estimated ± 2.4 pp, medium confidence |
| ScarfBench | 68.7% | estimated ± 4.5 pp, low confidence |
| BlenderWorld (text-only) | 70.9% | estimated ± 2.8 pp, medium confidence |
| Cookie-Bench | 79.2% | estimated ± 1.3 pp, medium confidence |
| STEMBench | 2.5% | estimated ± 0.2 pp, medium confidence |
| SimpleQA | 38.3% | estimated ± 4.3 pp, low confidence |
| Agents' Last Exam | 34.2% | estimated ± 9.7 pp, medium confidence |
| OpenHarmony Bench v1.0 | 58.9% | estimated ± 2.4 pp, high confidence |
| OpenHarmony Bench v1.2 | 63.8% | estimated ± 11.0 pp, low confidence |
| SEC-Bench Pro | 35.8% | estimated ± 5.6 pp, medium confidence |
| MRCRv2 | 85.9% | estimated ± 11.3 pp, low confidence |
| MMLongBench-Doc | 61.5% | estimated ± 2.6 pp, low confidence |
| SWE-Marathon | 43.7% | estimated ± 2.1 pp, medium confidence |
| BioBasicsBench | 34.5% | estimated ± 2.8 pp, medium confidence |
| LifeVal | 82.2% | estimated ± 2.3 pp, medium confidence |
| MMMU-Pro w/ Python | 77.6% | estimated ± 2.7 pp, medium confidence |
| ExploitBench Internal Port | 3.8% | estimated ± 1.4 pp, medium confidence |