NVIDIA releases a quantized FP4 local version of Qwen3.8-Flash-Next
AINVIDIA has published Qwen3.8-Flash-Next-FP4-Local on Hugging Face, a quantized version of Alibaba's Qwen3.8-Flash-Next model. The checkpoint is about 96 GiB and is validated for local hardware such as the RTX PRO 6000 Blackwell and DGX Spark.
Source details
NVIDIA · new models on Hugging Face · huggingface.coPublished · added here
Original title: nvidia/Qwen3.8-Flash-Next-FP4-Local