| Benchmark | o4-mini | Qwen3.5 397B A17B |
|---|---|---|
| AA Agentic Index | 36.1 | 53.3 |
| AA Intelligence | 26.1 | 34.3 |
| AA-LCR | 60 | 72.7 |
| AA-Omniscience | -35.7 | -30.8 |
| arena_vision | 1201 | 1249 |
| Artificial Analysis Coding Index | 25.6 | 48.2 |
| browsecomp | 51.5 | 69 |
| critpt | 0.6 | 1.7 |
| gdpval | 25.4 | 35.8 |
| GPQA Diamond | 81.4 | 89.3 |
| GSM8K | 0 | 66.7 |
| HLE | 16.5 | 29 |
| IFBench | 68.7 | 78.8 |
| mmlu_prox | 69.3 | 84.7 |
| MMMU | 81.6 | 85 |
| MMMU-Pro | 69.2 | 79 |
| multichallenge | 44.9 | 67.6 |
| OmniScience Accuracy | 24.8 | 30.8 |
| OmniScience Non-Hallucination | 19.5 | 17.3 |
| scicode | 46.5 | 42 |
| simpleqa_verified | 23.9 | 26 |
| strongreject | 96 | 99.4 |
| SWE-bench Verified | 68.1 | 76.4 |
| Terminal-Bench Hard | 15.2 | 40.9 |
| τ²-Bench Telecom (AA run) | 55.6 | 95.6 |
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.