| Benchmark | Claude 3.7 Sonnet | Qwen3.5 397B A17B |
|---|---|---|
| AA Agentic Index | 37 | 53.3 |
| AA Intelligence | 27.6 | 34.3 |
| AA-LCR | 62.3 | 72.7 |
| AA-Omniscience | -0.7 | -30.8 |
| arena_vision | 1195 | 1249 |
| Artificial Analysis Coding Index | 36.4 | 48.2 |
| critpt | 0.9 | 1.7 |
| gdpval | 27.4 | 35.8 |
| GPQA Diamond | 84.8 | 89.3 |
| HLE | 9.7 | 29 |
| hmmt_feb_2025 | 31.7 | 94.8 |
| IFBench | 48.3 | 78.8 |
| ifeval | 93.2 | 92.6 |
| mmmlu | 86.1 | 88.5 |
| MMMU | 75 | 85 |
| MMMU-Pro | 60.1 | 79 |
| multichallenge | 51.6 | 67.6 |
| OmniScience Accuracy | 28.2 | 30.8 |
| OmniScience Non-Hallucination | 60 | 17.3 |
| scicode | 40.3 | 42 |
| SWE-bench Verified | 70.3 | 76.4 |
| Terminal-Bench Hard | 21.2 | 40.9 |
| τ²-Bench Telecom (AA run) | 54.7 | 95.6 |
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.