| Benchmark | DeepSeek-R1 | Qwen3.7 Max |
|---|---|---|
| AA Agentic Index | 3.1 | 30.9 |
| AA Intelligence | 18.6 | 46.7 |
| AA-LCR | 56 | 74.7 |
| AA-Omniscience | -31.3 | 13.5 |
| Artificial Analysis Coding Index | 24.6 | 66 |
| critpt | 0.6 | 13.4 |
| gdpval | 1.5 | 38.5 |
| GPQA Diamond | 81 | 92.4 |
| HLE | 17.7 | 41.4 |
| IFBench | 39 | 80.5 |
| ifeval | 83.3 | 94.3 |
| longbench_v2 | 58.3 | 65.3 |
| mmlu_prox | 75.5 | 87 |
| mmlu_redux | 93.4 | 95 |
| MMLU-Pro | 85 | 89.6 |
| OmniScience Accuracy | 30.7 | 31.1 |
| OmniScience Non-Hallucination | 10.5 | 74.4 |
| scicode | 35.7 | 53.5 |
| simplebench | 40.8 | 70.4 |
| simpleqa_verified | 27.4 | 58.5 |
| SWE-bench Multilingual | 30.5 | 78.3 |
| SWE-bench Verified | 57.6 | 80.4 |
| TauBench V3 - Banking | 6.4 | 11.8 |
| Terminal-Bench 2.1 | 19.1 | 75 |
| Terminal-Bench Hard | 6.1 | 50.8 |
| τ²-Bench Telecom (AA run) | 11.4 | 94.7 |
| τ³-Bench | 6.4 | 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.