## Summary Normalizes Qwen model-family names before reasoning extraction so provider-prefixed Qwen models use the existing `<think>...</think>` fallback.