| Benchmark | o1 | Qwen3.5 27B |
|---|---|---|
| AA Agentic Index | 31.1 | 54.6 |
| AA Intelligence | 24 | 34.6 |
| AA-LCR | 63.3 | 72.3 |
| AA-Omniscience | -10.6 | -44 |
| arena_vision | 1193 | 1221 |
| Artificial Analysis Coding Index | 39.7 | 34.9 |
| critpt | 0.3 | 0.9 |
| gdpval | 11.5 | 33 |
| GPQA Diamond | 75.7 | 85.8 |
| HLE | 7.7 | 48.5 |
| IFBench | 70.3 | 76.5 |
| MathVista | 71.8 | 87.8 |
| mmmlu | 87.7 | 85.9 |
| MMMU | 77.6 | 82.3 |
| OmniScience Accuracy | 34.7 | 20.7 |
| OmniScience Non-Hallucination | 30.7 | 24.9 |
| scicode | 35.8 | 39.5 |
| SWE-bench Verified | 48.9 | 75 |
| Terminal-Bench Hard | 12.9 | 32.6 |
| τ²-Bench Telecom (AA run) | 62.6 | 93.9 |
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.