| Benchmark | GPT-5 | Qwen3.7 Plus Preview |
|---|---|---|
| AA Agentic Index | 49.7 | 20.8 |
| AA Intelligence | 35.3 | 39.4 |
| AA-LCR | 76.3 | 69 |
| AA-Omniscience | -8.7 | 2.4 |
| arena_vision | 1212 | 1263 |
| Artificial Analysis Coding Index | 38.9 | 55.9 |
| critpt | 5.7 | 9.1 |
| ERQA | 65.7 | 69.8 |
| gdpval | 32.5 | 22.2 |
| GPQA Diamond | 87.3 | 90.3 |
| HLE | 28.5 | 35.6 |
| IFBench | 73.1 | 79.1 |
| imo_answer_bench | 76 | 86 |
| LiveCodeBench v6 | 87 | 89.6 |
| mmlu_redux | 95.3 | 94.5 |
| MMLU-Pro | 87.5 | 88.5 |
| MMMU-Pro | 78.4 | 80.5 |
| ocrbench_v2 | 55.5 | 67.1 |
| OmniScience Accuracy | 40.3 | 22.5 |
| OmniScience Non-Hallucination | 21 | 74.5 |
| scicode | 43 | 51.3 |
| SWE-bench Multilingual | 55.3 | 75.8 |
| SWE-bench Pro | 41.8 | 57.6 |
| SWE-bench Verified | 74.9 | 77.7 |
| TauBench V3 - Banking | 22.1 | 17.9 |
| Terminal-Bench 2.0 | 35.2 | 70.3 |
| Terminal-Bench 2.1 | 35.2 | 61 |
| Terminal-Bench Hard | 37.9 | 47 |
| VideoMMMU | 84.6 | 85.4 |
| τ²-Bench Telecom (AA run) | 86.5 | 93 |
| τ³-Bench | 19.6 | 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.