| Benchmark | GPT-5.1 | Qwen3.7 Max |
|---|---|---|
| AA Agentic Index | 32.2 | 30.9 |
| AA Intelligence | 37.5 | 46.7 |
| AA-LCR | 76.7 | 74.7 |
| AA-Omniscience | 5.4 | 13.5 |
| arena_elo | 1439 | 1474 |
| arena_text_factuality | 1445 | 1479 |
| Artificial Analysis Coding Index | 49.4 | 66 |
| coding_arena_elo | 1421 | 1517 |
| critpt | 4.9 | 13.4 |
| gdpval | 25 | 38.5 |
| GPQA Diamond | 88.1 | 92.4 |
| HLE | 28.5 | 41.4 |
| HLE (with tools) | 42.7 | 53.5 |
| hmmt_nov_2025 | 91.7 | 95 |
| IFBench | 72.9 | 80.5 |
| LiveCodeBench v6 | 87 | 91.6 |
| MCP Atlas | 50.1 | 76.4 |
| MMLU-Pro | 87 | 89.6 |
| mmmlu | 91 | 90.3 |
| OmniScience Accuracy | 37.7 | 31.1 |
| OmniScience Non-Hallucination | 48.1 | 74.4 |
| scicode | 43.3 | 53.5 |
| simplebench | 53.2 | 70.4 |
| simpleqa_verified | 48.9 | 58.5 |
| SWE-bench Verified | 76.3 | 80.4 |
| TauBench V3 - Banking | 15.9 | 11.8 |
| Terminal-Bench 2.0 | 47.6 | 69.7 |
| Terminal-Bench 2.1 | 52.4 | 75 |
| Terminal-Bench Hard | 45.5 | 50.8 |
| τ²-Bench Telecom (AA run) | 81.9 | 94.7 |
| τ³-Bench | 14 | 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.