| Benchmark | Gemini 3.6 Flash | Qwen3.6 Plus |
|---|---|---|
| AA Agentic Index | 40.5 | 29 |
| AA Intelligence | 52 | 40.5 |
| AA-LCR | 79 | 72.3 |
| AA-Omniscience | 22.1 | 2.6 |
| aime_2026 | 96.7 | 95.3 |
| arena_elo | 1481 | 1444 |
| arena_text_factuality | 1470 | 1438 |
| Artificial Analysis Coding Index | 69.2 | 54.5 |
| coding_arena_elo | 1538 | 1460 |
| critpt | 10.6 | 2.9 |
| gdpval | 46.1 | 32 |
| GPQA Diamond | 92.8 | 90.4 |
| HLE | 40.8 | 28.8 |
| hmmt_feb_2026 | 89.4 | 87.8 |
| livebench_agentic_coding | 43.4 | 41.4 |
| livebench_coding | 77.9 | 78.2 |
| livebench_data_analysis | 63 | 69.9 |
| livebench_instruction_following | 75.4 | 58.3 |
| livebench_language | 83.9 | 75 |
| livebench_math | 86.4 | 83.7 |
| livebench_reasoning | 85.2 | 75.8 |
| MMMU-Pro | 83.2 | 78.8 |
| OmniScience Accuracy | 50 | 26.4 |
| OmniScience Non-Hallucination | 44.4 | 68 |
| OSWorld-Verified | 83 | 62.5 |
| scicode | 52.7 | 40.7 |
| simpleqa_verified | 68.7 | 49.1 |
| SWE-bench Pro | 58.7 | 56.6 |
| TauBench V3 - Banking | 29.9 | 20.8 |
| Terminal-Bench 2.1 | 78 | 61.4 |
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.