hugpadhugpad

Mia-AiLab / Qwen3.8-27B-DFlash2-EXL3-5.0bpw

Text Generation735M65210Updated Aug 25, 2026
Hugging Face

Downloads

652

Likes

10

Last updated

Aug 25, 2026

License

apache-2.0

EXL3 quantization of the DFlash2 draft model for Qwen/Qwen3.8-27B — a 1.93B block-diffusion speculative-decoding drafter (original bf16 release: incoai/Qwen3.8-27B-DFlash2, DFlash2 blog, GitHub).

Tasks & library

  • text generation
  • exllamav3

Tags

  • exllamav3
  • qwen3
  • exl3
  • draft-model
  • speculative-decoding
  • dflash2

Technical specifications

Parameters735M
Libraryexllamav3
LanguagesNot available
Base modelincoai/Qwen3.8-27B-DFlash2
Licenseapache-2.0