| Benchmark | Claude 3.7 Sonnet | DeepSeek-R1 |
|---|---|---|
| AA Agentic Index | 37 | 3.1 |
| AA Intelligence | 27.6 | 18.6 |
| AA-LCR | 62.3 | 56 |
| AA-Omniscience | -0.7 | -31.3 |
| aider_polyglot | 64.9 | 71.6 |
| AIME 2025 | 54.8 | 87.5 |
| AIR-Bench 2024 | 81.8 | 52.9 |
| anthropic_red_team | 99.7 | 99.1 |
| arc_agi_1 | 13.6 | 21.2 |
| ARC-AGI-2 | 0 | 1.3 |
| Artificial Analysis Coding Index | 36.4 | 24.6 |
| bbq | 92.1 | 96.6 |
| critpt | 0.9 | 0.6 |
| Fortress | 38 | 74.4 |
| gdpval | 27.4 | 1.5 |
| GPQA Diamond | 84.8 | 81 |
| harmbench | 84.3 | 54.6 |
| HLE | 9.7 | 17.7 |
| hmmt_feb_2025 | 31.7 | 76.7 |
| IFBench | 48.3 | 39 |
| ifeval | 93.2 | 83.3 |
| math_500_em | 96.2 | 98 |
| multichallenge | 51.6 | 45 |
| MultiNRC | 27.8 | 27.6 |
| OmniScience Accuracy | 28.2 | 30.7 |
| OmniScience Non-Hallucination | 60 | 10.5 |
| scicode | 40.3 | 35.7 |
| simple_safety_tests | 100 | 98.3 |
| simplebench | 46.4 | 40.8 |
| SWE-bench Verified | 70.3 | 57.6 |
| TAU-bench (airline) | 58.4 | 53.5 |
| TAU-bench (retail) | 81.2 | 63.9 |
| Terminal-Bench | 35.2 | 5.7 |
| Terminal-Bench Hard | 21.2 | 6.1 |
| usamo_2025 | 3.6 | 30.1 |
| xstest | 96.4 | 98.8 |
| τ²-Bench Telecom (AA run) | 54.7 | 11.4 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.