| Benchmark | Gemini 3.1 Pro | Qwen3.6 27B |
|---|---|---|
| AA Agentic Index | 23 | 27.5 |
| AA Intelligence | 47.7 | 37.7 |
| AA-LCR | 79 | 73.3 |
| AA-Omniscience | 32.9 | -20 |
| aime_2026 | 98.3 | 94.1 |
| Artificial Analysis Coding Index | 68.8 | 53.7 |
| baby_vision | 51.6 | 28.9 |
| charxiv_rq | 89.9 | 78.4 |
| critpt | 17.7 | 1.1 |
| DeepSWE 1.1 | 12 | 13.3 |
| gdpval | 23.3 | 32 |
| GPQA Diamond | 94.3 | 87.8 |
| HLE | 51.4 | 24 |
| hmmt_feb_2026 | 94.7 | 84.3 |
| hmmt_nov_2025 | 94.8 | 90.7 |
| IFBench | 77.1 | 69.1 |
| imo_answer_bench | 91 | 80.8 |
| livebench_agentic_coding | 44.1 | 39.3 |
| livebench_coding | 76.5 | 71.8 |
| livebench_data_analysis | 78.5 | 70.4 |
| livebench_instruction_following | 79.1 | 53.2 |
| livebench_language | 85.4 | 63.3 |
| livebench_math | 91 | 79.9 |
| livebench_reasoning | 84 | 70.3 |
| LiveCodeBench v6 | 91.7 | 83.9 |
| mathvision | 89.8 | 85.1 |
| MMLU-Pro | 91 | 86.2 |
| MMMU-Pro | 83 | 75.8 |
| nl2repo | 33.4 | 36.2 |
| OmniScience Accuracy | 55.3 | 19.6 |
| OmniScience Non-Hallucination | 50.1 | 50.7 |
| OSWorld-Verified | 76.2 | 63.9 |
| scicode | 59 | 39.8 |
| SWE-bench Multilingual | 76.9 | 71.3 |
| SWE-bench Pro | 54.2 | 53.5 |
| SWE-bench Verified | 80.6 | 77.2 |
| TauBench V3 - Banking | 21.4 | 16.7 |
| Terminal-Bench 2.0 | 68.5 | 59.3 |
| Terminal-Bench 2.1 | 74 | 60.7 |
| Terminal-Bench Hard | 68.5 | 34.8 |
| τ²-Bench Telecom (AA run) | 99.3 | 94.2 |
| τ³-Bench | 67.1 | 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.