| Benchmark | Gemini 3.6 Flash | Qwen3.7 Plus Preview |
|---|---|---|
| AA Agentic Index | 40.5 | 20.8 |
| AA Intelligence | 52 | 39.4 |
| AA-LCR | 79 | 69 |
| AA-Omniscience | 22.1 | 2.4 |
| arena_elo | 1481 | 1458 |
| arena_text_factuality | 1470 | 1454 |
| Artificial Analysis Coding Index | 69.2 | 55.9 |
| charxiv_rq | 85.2 | 85.9 |
| critpt | 10.6 | 9.1 |
| DeepSWE 1.1 | 49 | 14.2 |
| gdpval | 46.1 | 22.2 |
| GPQA Diamond | 92.8 | 90.3 |
| HLE | 40.8 | 35.6 |
| hmmt_feb_2026 | 89.4 | 92.9 |
| LVBench | 84.2 | 76.2 |
| MMMU-Pro | 83.2 | 80.5 |
| OmniScience Accuracy | 50 | 22.5 |
| OmniScience Non-Hallucination | 44.4 | 74.5 |
| OSWorld-Verified | 83 | 73.3 |
| scicode | 52.7 | 51.3 |
| SWE-bench Pro | 58.7 | 57.6 |
| TauBench V3 - Banking | 29.9 | 17.9 |
| Terminal-Bench 2.1 | 78 | 61 |
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.