hugpadhugpad

PrimeIntellect / DeepSeek-V4-Flash-0731-bf16

Text Generation284B10K0Updated Sep 14, 2026
Hugging Face

Downloads

10K

Likes

0

Last updated

Sep 14, 2026

License

mit

A bfloat16 re-export of deepseek-ai/DeepSeek-V4-Flash-0731. The weights are the published ones, dequantized: the dense linears from fp8 (e4m3 with ue8m0 scales over 128x128 blocks) and the routed experts from MXFP4, both to plain bfloat16. 43 layers, 114 safetensors shards.

Tasks & library

  • text generation
  • transformers

Tags

  • transformers
  • deepseek_v4
  • text-generation

Technical specifications

Parameters284B
Librarytransformers
LanguagesNot available
Base modeldeepseek-ai/DeepSeek-V4-Flash-0731
Licensemit