On August 28, 2026, NVIDIA released the model repository "nvidia/Muse-Glimmer-30B-NVFP4" on Hugging Face. This model is the NVFP4 quantized build of Muse-Glimmer-30B and is classified as an image-text-to-text model that accepts image and text inputs to produce text output. It has also been tagged as "conversational."
The base model is listed as meta-models/Muse-Glimmer-30B, and NVIDIA's Model Optimizer (ModelOpt) was used for the quantization. The file format is safetensors, and the license is Apache 2.0. The tags also include quantized, FP4, and 8-bit.
Source: nvidia/Muse-Glimmer-30B-NVFP4 (HF: NVIDIA, 2026-08-28)