SYLIN

Qwen3.5-4B-Qwen3.8-27B-Draft

by NVIDIA · Model · Language & code models

New open model on Hugging Face (text-generation): 4.21B parameters, 8.41 GB of weights, 2 community quantizations incl. GGUF

AccessOpen weightsNeeds~8.4 GB; smallest build ~2.5 GBDownload8.4 GBSize4.21B parametersLicenseNVIDIA Open Model Agreement

Evidence

  1. Nov 18
    NVIDIA has a 4B draft model for speculative decoding of Qwen3.8-27B

Cost to run