A methodical and thorough reasoner that handles complex multi-step problems with notable consistency. Particularly strong in structured tasks like coding, mathematics, and instruction-following, where it tends to stay on track without drifting. At 72B parameters it carries real depth, though it can be verbose when a concise answer would do.
| Benchmark | Score | Type | Recorded |
|---|---|---|---|
| IFEval | 86.4 | accuracy | 4d ago |
| GSM8K | 95.8 | accuracy | 5mo ago |