| Benchmark | Gemini 3.6 Flash | Qwen3.5 27B |
|---|---|---|
| AA Agentic Index | 40.5 | 54.6 |
| AA Intelligence | 52 | 34.6 |
| AA-LCR | 79 | 72.3 |
| AA-Omniscience | 22.1 | -44 |
| aime_2026 | 96.7 | 92.6 |
| Artificial Analysis Coding Index | 69.2 | 34.9 |
| charxiv_rq | 85.2 | 79.5 |
| coding_arena_elo | 1538 | 1357 |
| critpt | 10.6 | 0.9 |
| gdpval | 46.1 | 33 |
| GPQA Diamond | 92.8 | 85.8 |
| HLE | 40.8 | 48.5 |
| hmmt_feb_2026 | 89.4 | 84.3 |
| LVBench | 84.2 | 73.6 |
| MMMU-Pro | 83.2 | 75 |
| OmniScience Accuracy | 50 | 20.7 |
| OmniScience Non-Hallucination | 44.4 | 24.9 |
| OSWorld-Verified | 83 | 56.2 |
| scicode | 52.7 | 39.5 |
| SWE-bench Pro | 58.7 | 51.2 |
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.