hugpadhugpad

nerkyor / Qwen3.6-35B-A3B-DSV4Pro-Thinking-Distill-FP8

Agents37B441Updated Jun 28, 2026
Hugging Face

Downloads

44

Likes

1

Last updated

Jun 28, 2026

License

apache-2.0

The FP8 (block-128 e4m3) + native MTP build of Qwen3.6-35B-A3B-DSV4Pro-Thinking-Distill — for SGLang / vLLM serving. 35B-A3B is MoE (256 experts, 3B active); sparse activation makes it fast even single-stream.

Tasks & library

  • text generation

Tags

  • qwen3_5_moe
  • qwen3
  • moe
  • a3b
  • distillation
  • reasoning

Technical specifications

Parameters37B
LibraryNot available
Languageszh, en
Base modelMerkyor/Qwen3.6-35B-A3B-DSV4Pro-Thinking-Distill
Licenseapache-2.0