| Benchmark | Claude Sonnet 5 | DeepSeek-R1 |
|---|---|---|
| AA Agentic Index | 49.7 | 3.1 |
| AA Intelligence | 55.3 | 18.6 |
| AA-LCR | 77 | 56 |
| AA-Omniscience | 16.4 | -31.3 |
| Artificial Analysis Coding Index | 71.5 | 24.6 |
| browsecomp | 84.7 | 8.9 |
| critpt | 16.9 | 0.6 |
| gdpval | 54.8 | 1.5 |
| GPQA Diamond | 91.1 | 81 |
| HLE | 57.4 | 17.7 |
| OmniScience Accuracy | 40 | 30.7 |
| OmniScience Non-Hallucination | 60.6 | 10.5 |
| scicode | 53.6 | 35.7 |
| simplebench | 60.6 | 40.8 |
| simpleqa_verified | 25 | 27.4 |
| SWE-bench Multilingual | 78.3 | 30.5 |
| SWE-bench Verified | 85.2 | 57.6 |
| TauBench V3 - Banking | 37.3 | 6.4 |
| Terminal-Bench 2.1 | 80.5 | 19.1 |
| τ³-Bench | 28.2 | 6.4 |
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.