| Benchmark | Claude 3.7 Sonnet | Qwen3.6 27B |
|---|---|---|
| AA Agentic Index | 37 | 27.5 |
| AA Intelligence | 27.6 | 37.7 |
| AA-LCR | 62.3 | 73.3 |
| AA-Omniscience | -0.7 | -20 |
| Artificial Analysis Coding Index | 36.4 | 53.7 |
| critpt | 0.9 | 1.1 |
| gdpval | 27.4 | 32 |
| GPQA Diamond | 84.8 | 87.8 |
| HLE | 9.7 | 24 |
| hmmt_feb_2025 | 31.7 | 93.8 |
| IFBench | 48.3 | 69.1 |
| MMMU | 75 | 82.9 |
| MMMU-Pro | 60.1 | 75.8 |
| OmniScience Accuracy | 28.2 | 19.6 |
| OmniScience Non-Hallucination | 60 | 50.7 |
| scicode | 40.3 | 39.8 |
| SWE-bench Verified | 70.3 | 77.2 |
| Terminal-Bench Hard | 21.2 | 34.8 |
| τ²-Bench Telecom (AA run) | 54.7 | 94.2 |
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.