A mid-sized model that punches above its weight on reasoning and instruction-following tasks, with the standout trait of handling up to 1 million tokens of context — enough to process entire codebases or lengthy document collections in a single pass. It tends to be methodical and thorough, though like most models at this scale it can lose precision on highly complex multi-step problems. Multilingual capability is solid, reflecting Qwen's strong focus on Chinese and English.
| Benchmark | Score | Type | Recorded |
|---|---|---|---|
| IFEval | 84.1 | accuracy | 4d ago |