| Benchmark | Gemini 3.6 Flash | Qwen3.5 397B A17B |
|---|---|---|
| AA Agentic Index | 40.5 | 53.3 |
| AA Intelligence | 52 | 34.3 |
| AA-LCR | 79 | 72.7 |
| AA-Omniscience | 22.1 | -30.8 |
| aime_2026 | 96.7 | 94.2 |
| arena_elo | 1481 | 1442 |
| arena_text_factuality | 1470 | 1445 |
| Artificial Analysis Coding Index | 69.2 | 48.2 |
| charxiv_rq | 85.2 | 80.8 |
| coding_arena_elo | 1538 | 1399 |
| critpt | 10.6 | 1.7 |
| gdpval | 46.1 | 35.8 |
| GDPval-AA v2 | 1421 | 962 |
| GPQA Diamond | 92.8 | 89.3 |
| HLE | 40.8 | 29 |
| hmmt_feb_2026 | 89.4 | 87.9 |
| MMMU-Pro | 83.2 | 79 |
| OmniScience Accuracy | 50 | 30.8 |
| OmniScience Non-Hallucination | 44.4 | 17.3 |
| scicode | 52.7 | 42 |
| simpleqa_verified | 68.7 | 26 |
| SWE-bench Pro | 58.7 | 50.9 |
| TauBench V3 - Banking | 29.9 | 13.4 |
| Terminal-Bench 2.1 | 78 | 51.3 |
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.