| Benchmark | DeepSeek-V3 | Qwen3.7 Max |
|---|---|---|
| AA Agentic Index | 1.6 | 30.9 |
| AA Intelligence | 15.4 | 46.7 |
| AA-LCR | 41.3 | 74.7 |
| AA-Omniscience | -37.6 | 13.5 |
| Artificial Analysis Coding Index | 23 | 66 |
| critpt | 0 | 13.4 |
| gdpval | 0 | 38.5 |
| GPQA Diamond | 68.4 | 92.4 |
| HLE | 5.2 | 41.4 |
| IFBench | 41 | 80.5 |
| ifeval | 87.3 | 94.3 |
| livebench | 72.4 | 74.3 |
| LiveCodeBench v6 | 46.9 | 91.6 |
| longbench_v2 | 48.7 | 65.3 |
| mmlu_prox | 70.5 | 87 |
| mmlu_redux | 90.5 | 95 |
| MMLU-Pro | 81.2 | 89.6 |
| mmmlu | 79.4 | 90.3 |
| OmniScience Accuracy | 25.4 | 31.1 |
| OmniScience Non-Hallucination | 23.3 | 74.4 |
| scicode | 35.8 | 53.5 |
| simplebench | 27.2 | 70.4 |
| supergpqa | 53.7 | 73.6 |
| SWE-bench Verified | 42 | 80.4 |
| TauBench V3 - Banking | 4.7 | 11.8 |
| Terminal-Bench 2.1 | 16.9 | 75 |
| Terminal-Bench Hard | 15.2 | 50.8 |
| τ²-Bench Telecom (AA run) | 47.1 | 94.7 |
| τ³-Bench | 4.7 | 10.9 |
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.