Anthropic · model
Claude 3.5 Haiku benchmark scores
As of 2026-10-29, Claude 3.5 Haiku (Anthropic) has measured scores on 17 benchmarks and estimated scores on 69 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 40.8% | measured |
| AA-HLE | 3.6% | measured |
| AA-SciCode | 27.8% | estimated ± 3.9 pp, medium confidence |
| CritPt | 0.0% | measured |
| GDPval-AA | 0.0% | measured |
| GPQA Diamond | 42.1% | estimated ± 1.2 pp, high confidence |
| HLE | 0.6% | estimated ± 6.9 pp, low confidence |
| IFBench | 54.5% | estimated ± 6.6 pp, low confidence |
| AA-LCR | 27.3% | measured |
| SciCode | 27.6% | estimated ± 3.3 pp, medium confidence |
| BrowseComp | 0.0% | estimated ± 9.0 pp, low confidence |
| HLE w/ tools | 33.8% | estimated ± 5.3 pp, low confidence |
| IFEval | 85.9% | measured |
| LongBench v2 | 52.2% | estimated ± 5.8 pp, low confidence |
| τ²-bench results | 24.6% | measured |
| AA Coding Index | 23.6% | estimated ± 5.3 pp, medium confidence |
| AA-MMMU-Pro | 45.6% | measured |
| APEX-Agents | 0.0% | estimated ± 9.0 pp, low confidence |
| APEX-Agents-AA | 5.4% | estimated ± 9.4 pp, medium confidence |
| BioMysteryBench (human-difficult) | 0.0% | estimated ± 6.8 pp, low confidence |
| DeepSearchQA | 34.7% | estimated ± 9.6 pp, medium confidence |
| SWE-bench Verified | 40.6% | measured |
| BFCL v4 | 53.9% | estimated ± 2.7 pp, high confidence |
| CyberGym | 0.0% | estimated ± 6.0 pp, low confidence |
| JobBench | 0.0% | estimated ± 8.7 pp, low confidence |
| SWE-bench Pro | 14.7% | estimated ± 4.6 pp, medium confidence |
| WideResearch | 0.0% | estimated ± 7.8 pp, low confidence |
| AA Agentic Index | 0.0% | estimated ± 2.9 pp, high confidence |
| MMLU-Pro | 65.0% | measured |
| AA-IFBench | 42.8% | measured |
| AIME 2024 | 5.3% | measured |
| Artificial Analysis Intelligence Index | 8.9% | measured |
| BBH | 86.6% | measured |
| GPQA | 41.6% | measured |
| MMLU | 80.3% | measured |
| AA-Omniscience Accuracy | 13.2% | measured |
| MATH-500 | 49.0% | estimated ± 1.2 pp, medium confidence |
| MMMLU | 83.3% | estimated ± 0.7 pp, medium confidence |
| MMMU | 61.0% | estimated ± 2.7 pp, medium confidence |
| AA MMLU-Pro | 82.7% | estimated ± 0.6 pp, medium confidence |
| AIME 2025 | 20.0% | estimated ± 3.3 pp, medium confidence |
| Gert Labs | 31.4% | estimated ± 8.9 pp, medium confidence |
| AA Tau3 Banking | 0.0% | estimated ± 4.0 pp, medium confidence |
| ARC-AGI-1 | 45.2% | estimated ± 7.6 pp, low confidence |
| ARC-AGI-2 | 30.7% | estimated ± 14.6 pp, low confidence |
| MazeBench | 0.4% | estimated ± 5.4 pp, low confidence |
| OSWorld-Verified | 62.6% | estimated ± 8.1 pp, low confidence |
| ParseBench | 41.2% | estimated ± 5.5 pp, low confidence |
| Vals GPQA Diamond | 81.3% | estimated ± 5.3 pp, low confidence |
| Vals MMLU-Pro | 83.9% | estimated ± 2.1 pp, medium confidence |
| Vals SWE-bench | 22.2% | estimated ± 6.8 pp, low confidence |
| AA AutomationBench | 0.0% | estimated ± 14.8 pp, low confidence |
| CWE-Bench-AA | 0.0% | estimated ± 9.8 pp, low confidence |
| DeepsecBench-AA | 0.0% | estimated ± 8.6 pp, low confidence |
| GDP.pdf | 0.0% | estimated ± 5.2 pp, low confidence |
| AA ITBench | 35.7% | estimated ± 5.1 pp, low confidence |
| CWE-bench v1 | 0.0% | estimated ± 6.4 pp, low confidence |
| HLE w/o tools | 28.9% | estimated ± 4.6 pp, medium confidence |
| Toolathlon-Verified | 0.0% | estimated ± 3.1 pp, medium confidence |
| ARC-AGI-3 | 1.4% | estimated ± 7.7 pp, low confidence |
| Hermes Bench | 0.0% | estimated ± 2.9 pp, medium confidence |
| Hermes Index | 0.0% | estimated ± 8.7 pp, low confidence |
| CharXiv | 58.3% | estimated ± 6.2 pp, low confidence |
| AI2D_TEST | 88.0% | estimated ± 2.0 pp, low confidence |
| C-Eval | 88.3% | estimated ± 1.0 pp, low confidence |
| MCP Atlas | 56.5% | estimated ± 10.7 pp, low confidence |
| MMLU-Redux | 80.9% | estimated ± 0.6 pp, high confidence |
| MMMU-Pro | 45.5% | estimated ± 1.6 pp, high confidence |
| QwenClawBench | 55.9% | estimated ± 6.0 pp, low confidence |
| ScreenSpot Pro | 52.6% | estimated ± 11.2 pp, low confidence |
| SuperGPQA | 64.4% | estimated ± 9.3 pp, low confidence |
| Toolathlon | 27.0% | estimated ± 6.3 pp, low confidence |
| Video-MME (w/o subtitle) | 70.7% | estimated ± 1.7 pp, low confidence |
| HealthBench Hard | 0.0% | estimated ± 8.6 pp, low confidence |
| MMLU-Pro (Arcee) | 77.1% | estimated ± 4.0 pp, low confidence |
| React Native Evals | 71.4% | estimated ± 2.3 pp, low confidence |
| ResearchClawBench | 0.0% | estimated ± 2.2 pp, low confidence |
| BioMysteryBench (human-solvable) | 81.7% | estimated ± 0.6 pp, low confidence |
| DRACO | 61.5% | estimated ± 3.1 pp, low confidence |
| HealthBench (length-adjusted) | 53.9% | estimated ± 3.3 pp, medium confidence |
| HealthBench Professional (raw) | 50.1% | estimated ± 8.6 pp, low confidence |
| STEMBench | 2.4% | estimated ± 0.2 pp, medium confidence |
| SimpleQA | 8.8% | estimated ± 4.3 pp, low confidence |
| LiveCodeBench | 50.0% | estimated ± 14.6 pp, low confidence |
| MRCRv2 | 22.8% | estimated ± 11.3 pp, low confidence |
| MMLongBench-Doc | 55.5% | estimated ± 2.6 pp, low confidence |