| Benchmark | GPT-5.1 | Qwen3.7 Plus Preview |
|---|---|---|
| AA Agentic Index | 32.2 | 20.8 |
| AA Intelligence | 37.5 | 39.4 |
| AA-LCR | 76.7 | 69 |
| AA-Omniscience | 5.4 | 2.4 |
| arena_elo | 1439 | 1458 |
| arena_text_factuality | 1445 | 1454 |
| arena_vision | 1250 | 1263 |
| Artificial Analysis Coding Index | 49.4 | 55.9 |
| critpt | 4.9 | 9.1 |
| gdpval | 25 | 22.2 |
| GPQA Diamond | 88.1 | 90.3 |
| HLE | 28.5 | 35.6 |
| IFBench | 72.9 | 79.1 |
| LiveCodeBench v6 | 87 | 89.6 |
| MCP Atlas | 50.1 | 73.2 |
| MMLU-Pro | 87 | 88.5 |
| mmmlu | 91 | 89 |
| MMMU-Pro | 75.5 | 80.5 |
| OmniScience Accuracy | 37.7 | 22.5 |
| OmniScience Non-Hallucination | 48.1 | 74.5 |
| scicode | 43.3 | 51.3 |
| SWE-bench Verified | 76.3 | 77.7 |
| TauBench V3 - Banking | 15.9 | 17.9 |
| Terminal-Bench 2.0 | 47.6 | 70.3 |
| Terminal-Bench 2.1 | 52.4 | 61 |
| Terminal-Bench Hard | 45.5 | 47 |
| τ²-Bench Telecom (AA run) | 81.9 | 93 |
| τ³-Bench | 14 | 17.9 |
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.