Entity
DeepSeek-V3.1-Think
The thinking mode variant of DeepSeek-V3.1 with compressed reasoning chains for faster responses.
术语属性
- Model Type：Large Language Model (thinking mode)
- Performance Improvement：20-50% fewer output tokens with comparable performance to R1-0528
- Training Method：Chain-of-thought compression training