Anthropic · model
Claude Haiku 4.5 benchmark scores
As of 2026-11-22, Claude Haiku 4.5 (Anthropic) has measured scores on 18 benchmarks and estimated scores on 96 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 74.4% | estimated ± 3.6 pp, high confidence |
| AA-HLE | 9.3% | estimated ± 6.2 pp, medium confidence |
| AA-SciCode | 45.3% | estimated ± 3.8 pp, high confidence |
| Artificial Analysis Intelligence Index | 12.6% | estimated ± 5.8 pp, medium confidence |
| CritPt | 0.9% | estimated ± 5.3 pp, medium confidence |
| GDPval-AA | 45.5% | estimated ± 1.7 pp, low confidence |
| GPQA Diamond | 80.1% | estimated ± 1.8 pp, medium confidence |
| HLE | 25.9% | estimated ± 9.2 pp, low confidence |
| AA-Omniscience Accuracy | 22.2% | estimated ± 6.5 pp, medium confidence |
| SciCode | 44.5% | estimated ± 3.6 pp, high confidence |
| Terminal-Bench 2.1 | 45.5% | estimated ± 6.7 pp, low confidence |
| BrowseComp | 66.5% | estimated ± 6.2 pp, medium confidence |
| HLE w/ tools | 53.7% | estimated ± 7.5 pp, low confidence |
| AA Coding Index | 46.7% | estimated ± 5.3 pp, medium confidence |
| AA-MMMU-Pro | 64.0% | estimated ± 6.1 pp, medium confidence |
| APEX-Agents-AA | 19.9% | estimated ± 3.6 pp, medium confidence |
| DeepSearchQA | 75.2% | estimated ± 8.2 pp, low confidence |
| SWE-bench Verified | 73.3% | measured |
| CyberGym | 37.9% | estimated ± 11.4 pp, low confidence |
| JobBench | 16.0% | measured |
| SWE-bench Pro | 51.1% | estimated ± 4.5 pp, high confidence |
| WideResearch | 73.8% | estimated ± 1.2 pp, medium confidence |
| MMLU-Pro | 81.1% | estimated ± 4.8 pp, high confidence |
| FailureSensorIQ | 61.8% | measured |
| GPQA | 75.1% | estimated ± 4.5 pp, high confidence |
| MMLU | 88.4% | estimated ± 1.2 pp, high confidence |
| Vals GPQA Diamond | 72.2% | measured |
| Vals LiveCodeBench | 41.2% | measured |
| Vals MMLU-Pro | 78.7% | measured |
| Chess Puzzles | 8.0% | measured |
| AA MMLU-Pro | 83.5% | estimated ± 1.1 pp, medium confidence |
| Gert Labs | 33.6% | estimated ± 7.7 pp, medium confidence |
| AA-AnalystAgent | 31.4% | estimated ± 2.7 pp, low confidence |
| AA EnterpriseOps-Gym | 40.4% | estimated ± 1.2 pp, low confidence |
| AA Terminal-Bench 2.1 | 79.4% | estimated ± 2.5 pp, medium confidence |
| ApprenticeBench | 0.1% | estimated ± 6.0 pp, low confidence |
| ARC-AGI-1 | 49.3% | estimated ± 6.5 pp, low confidence |
| ARC-AGI-2 | 0.0% | estimated ± 13.8 pp, low confidence |
| CursorBench 3.1 | 40.4% | estimated ± 5.8 pp, low confidence |
| CursorBench 3.2 | 61.5% | estimated ± 3.3 pp, medium confidence |
| EBR-bench | 0.0% | estimated ± 9.1 pp, low confidence |
| FrontierCode 1.1 Main | 15.7% | estimated ± 4.2 pp, low confidence |
| FrontierSWE v2 | 5.5% | estimated ± 9.5 pp, low confidence |
| Furniture Assembly | 24.7% | estimated ± 14.1 pp, low confidence |
| MaxxBench | 32.8% | estimated ± 10.3 pp, low confidence |
| MazeBench | 0.0% | measured |
| Mystery Game Puzzles | 8.7% | estimated ± 2.2 pp, low confidence |
| NYT Connections | 14.3% | measured |
| OfficeQA Pro | 47.1% | estimated ± 7.0 pp, low confidence |
| OSWorld-Verified | 60.8% | estimated ± 8.4 pp, low confidence |
| ParseBench-1 | 49.3% | measured |
| Terminal-Bench 4.1 | 6.1% | estimated ± 14.9 pp, low confidence |
| Terminal-Bench Hard | 44.0% | estimated ± 3.9 pp, medium confidence |
| Vals SWE-bench | 66.6% | measured |
| Terminal-Bench 2.1 (Vals AI) | 43.8% | measured |
| VulcanBench v3 | 76.2% | measured |
| CWE-Bench-AA | 44.0% | estimated ± 10.6 pp, low confidence |
| DeepsecBench-AA | 21.9% | estimated ± 10.3 pp, low confidence |
| Bug Hunt Bench | 19.0% | estimated ± 8.6 pp, low confidence |
| CursorBench 4.0 | 30.8% | estimated ± 3.4 pp, medium confidence |
| CWE-bench v1 | 57.4% | estimated ± 5.1 pp, medium confidence |
| DeepSWE | 56.3% | estimated ± 7.2 pp, low confidence |
| DRACO | 86.9% | estimated ± 0.7 pp, low confidence |
| FrontierPhysics | 0.0% | estimated ± 6.1 pp, low confidence |
| HLE w/o tools | 24.6% | estimated ± 4.9 pp, high confidence |
| OSWorld 2.1 | 66.4% | estimated ± 6.1 pp, medium confidence |
| OSWorld 2.1 Offline | 15.7% | measured |
| PostTrainBench v1.1 | 12.5% | estimated ± 3.7 pp, medium confidence |
| SpireClimb | 0.0% | estimated ± 10.3 pp, low confidence |
| StealthMark | 30.2% | estimated ± 9.7 pp, low confidence |
| SWE-bench Multimodal | 0.0% | estimated ± 5.6 pp, low confidence |
| Toolathlon-Verified | 73.4% | estimated ± 2.6 pp, high confidence |
| AA AutomationBench | 52.8% | estimated ± 12.8 pp, low confidence |
| GDP.pdf | 16.4% | estimated ± 4.1 pp, low confidence |
| AA Terminal-Bench 4.0 | 0.0% | estimated ± 13.7 pp, low confidence |
| CWE-bench | 53.9% | estimated ± 7.9 pp, low confidence |
| MKA | 37.3% | estimated ± 3.4 pp, medium confidence |
| SkillsBench official leaderboard (BenchFlow) | 30.1% | measured |
| BinaryAudit | 10.5% | measured |
| CompileBench | 68.9% | measured |
| OTelBench | 5.8% | measured |
| Vibe Code Bench | 0.6% | estimated ± 8.4 pp, medium confidence |
| FACTS Multimodal | 31.8% | estimated ± 5.6 pp, low confidence |
| Hermes Index | 37.2% | estimated ± 3.1 pp, high confidence |
| ParseBench-2 | 59.8% | estimated ± 4.3 pp, high confidence |
| CharXiv | 79.8% | estimated ± 5.6 pp, low confidence |
| ExploitBench | 7.6% | estimated ± 13.4 pp, low confidence |
| ExploitGym | 11.9% | estimated ± 6.3 pp, medium confidence |
| Claw-Eval | 54.1% | estimated ± 4.5 pp, medium confidence |
| MCP Atlas | 38.1% | estimated ± 8.6 pp, medium confidence |
| MMMU-Pro | 71.0% | estimated ± 6.9 pp, medium confidence |
| NL2Repo | 38.9% | estimated ± 4.7 pp, medium confidence |
| PostTrainBench v1 | 7.8% | estimated ± 3.6 pp, medium confidence |
| ScreenSpot Pro | 72.6% | estimated ± 10.1 pp, low confidence |
| Terminal-Bench 2.0 | 53.4% | estimated ± 5.8 pp, medium confidence |
| Toolathlon | 29.7% | estimated ± 7.4 pp, low confidence |
| AA LiveCodeBench | 85.4% | estimated ± 2.6 pp, high confidence |
| React Native Evals | 76.6% | estimated ± 2.7 pp, medium confidence |
| ResearchClawBench | 14.9% | estimated ± 2.6 pp, low confidence |
| SWE-Rebench | 57.0% | estimated ± 3.6 pp, medium confidence |
| PostTrainBench | 10.3% | estimated ± 3.7 pp, medium confidence |
| DAYJOB | 0.4% | estimated ± 6.3 pp, low confidence |
| DAYJOB Healthcare | 0.1% | estimated ± 6.1 pp, low confidence |
| Finance Agent v2 | 56.1% | estimated ± 3.8 pp, medium confidence |
| ScarfBench | 38.2% | estimated ± 12.3 pp, low confidence |
| BlenderWorld (text-only) | 70.6% | estimated ± 4.0 pp, low confidence |
| Cookie-Bench | 73.4% | estimated ± 2.5 pp, low confidence |
| LiveCodeBench | 75.1% | estimated ± 14.6 pp, low confidence |
| Agents' Last Exam | 35.4% | estimated ± 11.0 pp, low confidence |
| OpenHarmony Bench v1.0 | 50.1% | estimated ± 3.2 pp, medium confidence |
| SEC-Bench Pro | 13.4% | estimated ± 12.3 pp, low confidence |
| LifeVal | 75.2% | estimated ± 3.8 pp, medium confidence |
| MMMU-Pro w/ Python | 74.2% | estimated ± 2.7 pp, medium confidence |
| ExploitBench Internal Port | 0.0% | estimated ± 1.4 pp, low confidence |