| Benchmark | Claude 3.7 Sonnet | Qwen3.5 27B |
|---|---|---|
| AA Agentic Index | 37 | 54.6 |
| AA Intelligence | 27.6 | 34.6 |
| AA-LCR | 62.3 | 72.3 |
| AA-Omniscience | -0.7 | -44 |
| arena_vision | 1195 | 1221 |
| Artificial Analysis Coding Index | 36.4 | 34.9 |
| critpt | 0.9 | 0.9 |
| gdpval | 27.4 | 33 |
| GPQA Diamond | 84.8 | 85.8 |
| HLE | 9.7 | 48.5 |
| hmmt_feb_2025 | 31.7 | 92 |
| IFBench | 48.3 | 76.5 |
| ifeval | 93.2 | 95 |
| mmmlu | 86.1 | 85.9 |
| MMMU | 75 | 82.3 |
| MMMU-Pro | 60.1 | 75 |
| multichallenge | 51.6 | 60.8 |
| OmniScience Accuracy | 28.2 | 20.7 |
| OmniScience Non-Hallucination | 60 | 24.9 |
| scicode | 40.3 | 39.5 |
| SWE-bench Verified | 70.3 | 75 |
| Terminal-Bench Hard | 21.2 | 32.6 |
| τ²-Bench Telecom (AA run) | 54.7 | 93.9 |
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.