NVIDIA releases GLM-5.3-Flash-FP4-Local, a quantized version of ZAI's GLM-5.3-Flash
AINVIDIA releases GLM-5.3-Flash-FP4-Local on Hugging Face, a quantized version of ZAI's GLM-5.3-Flash made with quantization-aware distillation. The model card says it needs about 166 GiB of vRAM and requires community vLLM containers until upstream vLLM supports its NVFP4 scale format.
Source details
NVIDIA · new models on Hugging Face · huggingface.coPublished · added here
Original title: nvidia/GLM-5.3-Flash-FP4-Local