DeepSeek-R1-FP4: FP4-optimized version of DeepSeek-R1 inference 25x faster
Comprehensive Introduction DeepSeek-R1-FP4 is a quantified language model open-sourced and optimized by NVIDIA, developed based on DeepSeek-R1 from DeepSeek AI. It is developed based on DeepSeek-R1 for DeepSeek AI. It is optimized by the TensorRT Model Opt...


































































































