| Benchmark | GPT-5.1 | Qwen3.6 27B |
|---|---|---|
| AA Agentic Index | 32.2 | 27.5 |
| AA Intelligence | 37.5 | 37.7 |
| AA-LCR | 76.7 | 73.3 |
| AA-Omniscience | 5.4 | -20 |
| Artificial Analysis Coding Index | 49.4 | 53.7 |
| critpt | 4.9 | 1.1 |
| gdpval | 25 | 32 |
| GPQA Diamond | 88.1 | 87.8 |
| HLE | 28.5 | 24 |
| hmmt_feb_2025 | 96.3 | 93.8 |
| hmmt_nov_2025 | 91.7 | 90.7 |
| IFBench | 72.9 | 69.1 |
| LiveCodeBench v6 | 87 | 83.9 |
| MMLU-Pro | 87 | 86.2 |
| MMMU | 85.4 | 82.9 |
| MMMU-Pro | 75.5 | 75.8 |
| OmniScience Accuracy | 37.7 | 19.6 |
| OmniScience Non-Hallucination | 48.1 | 50.7 |
| scicode | 43.3 | 39.8 |
| SWE-bench Verified | 76.3 | 77.2 |
| TauBench V3 - Banking | 15.9 | 16.7 |
| Terminal-Bench 2.0 | 47.6 | 59.3 |
| Terminal-Bench 2.1 | 52.4 | 60.7 |
| Terminal-Bench Hard | 45.5 | 34.8 |
| τ²-Bench Telecom (AA run) | 81.9 | 94.2 |
| τ³-Bench | 14 | 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.