| Benchmark | Claude Sonnet 5 | Qwen3.6 35B A3B |
|---|---|---|
| AA Agentic Index | 49.7 | 21.6 |
| AA Intelligence | 55.3 | 32.1 |
| AA-LCR | 77 | 66.7 |
| AA-Omniscience | 16.4 | -22.2 |
| Artificial Analysis Coding Index | 71.5 | 41.9 |
| browsecomp | 84.7 | 48.7 |
| charxiv_rq | 77 | 78 |
| critpt | 16.9 | 0.3 |
| gdpval | 54.8 | 27.8 |
| GDPval-AA v2 | 1618 | 1015 |
| GPQA Diamond | 91.1 | 86 |
| HLE | 57.4 | 22.2 |
| LVBench | 68.5 | 71.4 |
| MMMU-Pro | 77.3 | 75.3 |
| OmniScience Accuracy | 40 | 18.8 |
| OmniScience Non-Hallucination | 60.6 | 49.5 |
| scicode | 53.6 | 35.8 |
| SWE-bench Multilingual | 78.3 | 67.2 |
| SWE-bench Pro | 63.2 | 49.5 |
| SWE-bench Verified | 85.2 | 73.4 |
| TauBench V3 - Banking | 37.3 | 9.3 |
| Terminal-Bench 2.0 | 80.4 | 51.5 |
| Terminal-Bench 2.1 | 80.5 | 44.9 |
| toolathlon | 54.3 | 26.9 |
| τ³-Bench | 28.2 | 67.2 |
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.