| Benchmark | Claude 3.7 Sonnet | Qwen3.6 35B A3B |
|---|---|---|
| AA Agentic Index | 37 | 21.6 |
| AA Intelligence | 27.6 | 32.1 |
| AA-LCR | 62.3 | 66.7 |
| AA-Omniscience | -0.7 | -22.2 |
| Artificial Analysis Coding Index | 36.4 | 41.9 |
| critpt | 0.9 | 0.3 |
| gdpval | 27.4 | 27.8 |
| GPQA Diamond | 84.8 | 86 |
| HLE | 9.7 | 22.2 |
| hmmt_feb_2025 | 31.7 | 90.7 |
| IFBench | 48.3 | 64.4 |
| MMMU | 75 | 81.7 |
| MMMU-Pro | 60.1 | 75.3 |
| OmniScience Accuracy | 28.2 | 18.8 |
| OmniScience Non-Hallucination | 60 | 49.5 |
| scicode | 40.3 | 35.8 |
| SWE-bench Verified | 70.3 | 73.4 |
| Terminal-Bench Hard | 21.2 | 34.8 |
| τ²-Bench Telecom (AA run) | 54.7 | 95.3 |
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.