| Benchmark | GPT-5.1 | Qwen3.6 35B A3B |
|---|---|---|
| AA Agentic Index | 32.2 | 21.6 |
| AA Intelligence | 37.5 | 32.1 |
| AA-LCR | 76.7 | 66.7 |
| AA-Omniscience | 5.4 | -22.2 |
| Artificial Analysis Coding Index | 49.4 | 41.9 |
| browsecomp | 50.8 | 48.7 |
| critpt | 4.9 | 0.3 |
| gdpval | 25 | 27.8 |
| GPQA Diamond | 88.1 | 86 |
| HLE | 28.5 | 22.2 |
| hmmt_feb_2025 | 96.3 | 90.7 |
| hmmt_nov_2025 | 91.7 | 89.1 |
| IFBench | 72.9 | 64.4 |
| LiveCodeBench v6 | 87 | 80.4 |
| MCP Atlas | 50.1 | 62.8 |
| MMLU-Pro | 87 | 85.6 |
| MMMU | 85.4 | 81.7 |
| MMMU-Pro | 75.5 | 75.3 |
| OmniScience Accuracy | 37.7 | 18.8 |
| OmniScience Non-Hallucination | 48.1 | 49.5 |
| scicode | 43.3 | 35.8 |
| SWE-bench Verified | 76.3 | 73.4 |
| TauBench V3 - Banking | 15.9 | 9.3 |
| Terminal-Bench 2.0 | 47.6 | 51.5 |
| Terminal-Bench 2.1 | 52.4 | 44.9 |
| Terminal-Bench Hard | 45.5 | 34.8 |
| τ²-Bench Telecom (AA run) | 81.9 | 95.3 |
| τ³-Bench | 14 | 67.2 |
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.