| Benchmark | GPT-5 | Qwen3.7 Max |
|---|---|---|
| AA Agentic Index | 49.7 | 30.9 |
| AA Intelligence | 35.3 | 46.7 |
| AA-LCR | 76.3 | 74.7 |
| AA-Omniscience | -8.7 | 13.5 |
| Artificial Analysis Coding Index | 38.9 | 66 |
| coding_arena_elo | 1419 | 1517 |
| critpt | 5.7 | 13.4 |
| gdpval | 32.5 | 38.5 |
| GPQA Diamond | 87.3 | 92.4 |
| HealthBench | 67.2 | 54.5 |
| HLE | 28.5 | 41.4 |
| HLE (with tools) | 35.2 | 53.5 |
| hmmt_nov_2025 | 89.2 | 95 |
| IFBench | 73.1 | 80.5 |
| imo_answer_bench | 76 | 90 |
| LiveCodeBench v6 | 87 | 91.6 |
| mmlu_redux | 95.3 | 95 |
| MMLU-Pro | 87.5 | 89.6 |
| OmniScience Accuracy | 40.3 | 31.1 |
| OmniScience Non-Hallucination | 21 | 74.4 |
| scicode | 43 | 53.5 |
| simplebench | 56.7 | 70.4 |
| simpleqa_verified | 50.6 | 58.5 |
| SWE-bench Multilingual | 55.3 | 78.3 |
| SWE-bench Pro | 41.8 | 60.6 |
| SWE-bench Verified | 74.9 | 80.4 |
| TauBench V3 - Banking | 22.1 | 11.8 |
| Terminal-Bench 2.0 | 35.2 | 69.7 |
| Terminal-Bench 2.1 | 35.2 | 75 |
| Terminal-Bench Hard | 37.9 | 50.8 |
| τ²-Bench Telecom (AA run) | 86.5 | 94.7 |
| τ³-Bench | 19.6 | 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.