| Benchmark | Gemini 3.5 Flash | Qwen3.6 27B |
|---|---|---|
| AA Agentic Index | 70.4 | 27.5 |
| AA Intelligence | 52 | 37.7 |
| AA-LCR | 81 | 73.3 |
| AA-Omniscience | 21.2 | -20 |
| aime_2026 | 95 | 94.1 |
| Artificial Analysis Coding Index | 70.1 | 53.7 |
| charxiv_rq | 84.2 | 78.4 |
| critpt | 13.1 | 1.1 |
| DeepSWE 1.1 | 37 | 13.3 |
| gdpval | 57.8 | 32 |
| GPQA Diamond | 92.2 | 87.8 |
| HLE | 42.7 | 24 |
| hmmt_feb_2026 | 95.5 | 84.3 |
| IFBench | 76.3 | 69.1 |
| livebench_agentic_coding | 49 | 39.3 |
| livebench_coding | 78.2 | 71.8 |
| livebench_data_analysis | 64.9 | 70.4 |
| livebench_instruction_following | 75.6 | 53.2 |
| livebench_language | 84.6 | 63.3 |
| livebench_math | 88.2 | 79.9 |
| livebench_reasoning | 82 | 70.3 |
| MMMU-Pro | 84.3 | 75.8 |
| OmniScience Accuracy | 51.4 | 19.6 |
| OmniScience Non-Hallucination | 38.2 | 50.7 |
| OSWorld-Verified | 78.4 | 63.9 |
| scicode | 53.1 | 39.8 |
| SWE-bench Pro | 55.1 | 53.5 |
| TauBench V3 - Banking | 32.2 | 16.7 |
| Terminal-Bench 2.0 | 76.2 | 59.3 |
| Terminal-Bench 2.1 | 78.7 | 60.7 |
| Terminal-Bench Hard | 46.2 | 34.8 |
| τ²-Bench Telecom (AA run) | 95.6 | 94.2 |
| τ³-Bench | 25.4 | 15.3 |
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.