| Benchmark | Qwen3.7 Max | Qwen3 VL 4B (Reasoning) |
|---|---|---|
| AA Agentic Index | 30.9 | 14.4 |
| AA Intelligence | 46.7 | 7.7 |
| AA-LCR | 74.7 | 23 |
| AA-Omniscience | 13.5 | -68.4 |
| Artificial Analysis Coding Index | 66 | 6.7 |
| critpt | 13.4 | 0 |
| gdpval | 38.5 | 13.8 |
| GPQA Diamond | 92.4 | 64.1 |
| HLE | 41.4 | 4.6 |
| IFBench | 80.5 | 36.6 |
| ifeval | 94.3 | 82.6 |
| include | 86.2 | 64.6 |
| LiveCodeBench v6 | 91.6 | 51.3 |
| mmlu_prox | 87 | 65 |
| mmlu_redux | 95 | 86 |
| MMLU-Pro | 89.6 | 73.6 |
| OmniScience Accuracy | 31.1 | 12.2 |
| OmniScience Non-Hallucination | 74.4 | 8.2 |
| scicode | 53.5 | 17.1 |
| supergpqa | 73.6 | 46.8 |
| Terminal-Bench Hard | 50.8 | 1.5 |
| τ²-Bench Telecom (AA run) | 94.7 | 15.5 |
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.