| Benchmark | Gemini 3.5 Flash | Qwen3.6 35B A3B |
|---|---|---|
| AA Agentic Index | 70.4 | 21.6 |
| AA Intelligence | 52 | 32.1 |
| AA-LCR | 81 | 66.7 |
| AA-Omniscience | 21.2 | -22.2 |
| aime_2026 | 95 | 92.7 |
| Artificial Analysis Coding Index | 70.1 | 41.9 |
| charxiv_rq | 84.2 | 78 |
| critpt | 13.1 | 0.3 |
| gdpval | 57.8 | 27.8 |
| GDPval-AA v2 | 1357 | 1015 |
| GPQA Diamond | 92.2 | 86 |
| HLE | 42.7 | 22.2 |
| hmmt_feb_2026 | 95.5 | 83.6 |
| IFBench | 76.3 | 64.4 |
| MCP Atlas | 83.6 | 62.8 |
| MMMU-Pro | 84.3 | 75.3 |
| OmniScience Accuracy | 51.4 | 18.8 |
| OmniScience Non-Hallucination | 38.2 | 49.5 |
| scicode | 53.1 | 35.8 |
| SWE-bench Pro | 55.1 | 49.5 |
| TauBench V3 - Banking | 32.2 | 9.3 |
| Terminal-Bench 2.0 | 76.2 | 51.5 |
| Terminal-Bench 2.1 | 78.7 | 44.9 |
| Terminal-Bench Hard | 46.2 | 34.8 |
| toolathlon | 56.5 | 26.9 |
| τ²-Bench Telecom (AA run) | 95.6 | 95.3 |
| τ³-Bench | 25.4 | 67.2 |
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.