| Benchmark | Gemini 2.5 Flash | Qwen3.6 Plus |
|---|---|---|
| AA Agentic Index | 18.8 | 29 |
| AA Intelligence | 17.7 | 40.5 |
| AA-LCR | 65.7 | 72.3 |
| AA-Omniscience | -29.8 | 2.6 |
| arena_text_factuality | 1437 | 1438 |
| Artificial Analysis Coding Index | 22.2 | 54.5 |
| critpt | 1.4 | 2.9 |
| frontiermath_tier_4 | 4.2 | 8.3 |
| gdpval | 11.9 | 32 |
| GPQA Diamond | 82.8 | 90.4 |
| HLE | 12.1 | 28.8 |
| IFBench | 50.3 | 75.2 |
| MMMU | 79.7 | 86 |
| MMMU-Pro | 69.1 | 78.8 |
| OmniScience Accuracy | 26.1 | 26.4 |
| OmniScience Non-Hallucination | 24.7 | 68 |
| scicode | 35.9 | 40.7 |
| SWE-bench Verified | 60.4 | 78.8 |
| Terminal-Bench Hard | 13.6 | 43.9 |
| τ²-Bench Telecom (AA run) | 31.6 | 97.7 |
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.