| Benchmark | GPT-4o | Qwen3.7 Max |
|---|---|---|
| AA Agentic Index | 9.7 | 30.9 |
| AA Intelligence | 11.2 | 46.7 |
| AA-LCR | 55 | 74.7 |
| AA-Omniscience | -10.5 | 13.5 |
| Artificial Analysis Coding Index | 24.2 | 66 |
| critpt | 0 | 13.4 |
| gdpval | 0 | 38.5 |
| GPQA Diamond | 70.1 | 92.4 |
| HLE | 5.3 | 41.4 |
| IFBench | 36 | 80.5 |
| ifeval | 84.3 | 94.3 |
| LiveCodeBench v6 | 30.9 | 91.6 |
| longbench_v2 | 48.1 | 65.3 |
| mmlu_prox | 61.1 | 87 |
| mmlu_redux | 88 | 95 |
| MMLU-Pro | 74.7 | 89.6 |
| mmmlu | 81.4 | 90.3 |
| OmniScience Accuracy | 23.7 | 31.1 |
| OmniScience Non-Hallucination | 62.1 | 74.4 |
| scicode | 33.4 | 53.5 |
| simplebench | 17.8 | 70.4 |
| supergpqa | 42.4 | 73.6 |
| SWE-bench Verified | 38.8 | 80.4 |
| Terminal-Bench Hard | 8.3 | 50.8 |
| τ²-Bench Telecom (AA run) | 28.9 | 94.7 |
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.