| Benchmark | Claude Sonnet 5 | Qwen3 VL 4B (Reasoning) |
|---|---|---|
| AA Agentic Index | 49.7 | 14.4 |
| AA Intelligence | 55.3 | 7.7 |
| AA-LCR | 77 | 23 |
| AA-Omniscience | 16.4 | -68.4 |
| Artificial Analysis Coding Index | 71.5 | 6.7 |
| critpt | 16.9 | 0 |
| gdpval | 54.8 | 13.8 |
| GPQA Diamond | 91.1 | 64.1 |
| HLE | 57.4 | 4.6 |
| LVBench | 68.5 | 53.5 |
| MMMU-Pro | 77.3 | 57 |
| OmniScience Accuracy | 40 | 12.2 |
| OmniScience Non-Hallucination | 60.6 | 8.2 |
| OSWorld-Verified | 81.2 | 31.4 |
| scicode | 53.6 | 17.1 |
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.