| Benchmark | Gemini 2.5 Pro | Qwen3.7 Plus Preview |
|---|---|---|
| AA Agentic Index | 7.2 | 20.8 |
| AA Intelligence | 27 | 39.4 |
| AA-LCR | 66 | 69 |
| AA-Omniscience | -14.3 | 2.4 |
| arena_elo | 1446 | 1458 |
| arena_text_factuality | 1462 | 1454 |
| arena_vision | 1246 | 1263 |
| Artificial Analysis Coding Index | 46.7 | 55.9 |
| critpt | 2.6 | 9.1 |
| gdpval | 8.5 | 22.2 |
| GPQA Diamond | 84.4 | 90.3 |
| HLE | 22.5 | 35.6 |
| IFBench | 49 | 79.1 |
| mathvision | 73.3 | 90.3 |
| MMLU-Pro | 86 | 88.5 |
| MMMU-Pro | 74.9 | 80.5 |
| ocrbench_v2 | 59.3 | 67.1 |
| OmniScience Accuracy | 39 | 22.5 |
| OmniScience Non-Hallucination | 12.6 | 74.5 |
| scicode | 43 | 51.3 |
| SWE-bench Verified | 67.2 | 77.7 |
| TauBench V3 - Banking | 9.7 | 17.9 |
| Terminal-Bench 2.1 | 28.5 | 61 |
| Terminal-Bench Hard | 26.5 | 47 |
| Video-MME | 84.8 | 88 |
| τ²-Bench Telecom (AA run) | 54.1 | 93 |
| τ³-Bench | 9.3 | 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.