| Benchmark | GPT-5.6 Sol | Qwen3.6 27B |
|---|---|---|
| AA Agentic Index | 57.8 | 27.5 |
| AA Intelligence | 61 | 37.7 |
| AA-LCR | 77.7 | 73.3 |
| AA-Omniscience | 22 | -20 |
| Agents' Last Exam | 52.7 | 27.3 |
| aime_2026 | 99.9 | 94.1 |
| Artificial Analysis Coding Index | 78.3 | 53.7 |
| charxiv_rq | 87.8 | 78.4 |
| critpt | 32.3 | 1.1 |
| DeepSWE 1.1 | 73 | 13.3 |
| gdpval | 61.1 | 32 |
| GPQA Diamond | 94.6 | 87.8 |
| HLE | 49.5 | 24 |
| IFBench | 72.7 | 69.1 |
| JobBench | 45.4 | 21.8 |
| livebench_agentic_coding | 56.2 | 39.3 |
| livebench_coding | 83.9 | 71.8 |
| livebench_data_analysis | 79.8 | 70.4 |
| livebench_instruction_following | 71.8 | 53.2 |
| livebench_language | 87.7 | 63.3 |
| livebench_math | 96.2 | 79.9 |
| livebench_reasoning | 91.7 | 70.3 |
| mathvision | 95.8 | 85.1 |
| MMMU-Pro | 84.6 | 75.8 |
| OmniScience Accuracy | 59.4 | 19.6 |
| OmniScience Non-Hallucination | 10.6 | 50.7 |
| OSWorld-Verified | 83 | 63.9 |
| scicode | 56.9 | 39.8 |
| SWE-bench Pro | 64.6 | 53.5 |
| TauBench V3 - Banking | 44.3 | 16.7 |
| Terminal-Bench 2.1 | 89.5 | 60.7 |
| Terminal-Bench Hard | 65.9 | 34.8 |
| τ²-Bench Telecom (AA run) | 85.1 | 94.2 |
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.