NVIDIA 发布 Muse-Glimmer-30B NVFP4 量化模型
原标题：nvidia/Muse-Glimmer-30B-NVFP4
AI 摘要
NVIDIA 发布了 Muse-Glimmer-30B 的 NVFP4 量化版本，该模型基于 Meta Superintelligence Lab 的 Muse-Glimmer-30B 稠密因果 Transformer，带有专用感知编码器，支持文本与图像交错输入，面向本地智能体任务。量化使用 NVIDIA Model Optimizer v0.46.0，采用 Apache 2.0 许可，支持 vLLM 和 SGLang 运行时，并适配 NVIDIA Blackwell B200/B300 及 DGX Spark 硬件。
正文节选
--- pipeline_tag: image-text-to-text base_model: - meta-models/Muse-Glimmer-30B license: apache-2.0 library_name: Model Optimizer tags: - nvidia - ModelOpt - Muse-Glimmer - quantized - FP4 - fp4 --- # Model Overview ## Description: The NVIDIA Muse-Glimmer-30B NVFP4 model is the quantized version of Meta Superintelligence Lab's Muse-Glimmer-30B model, which is an auto-regressive language model that uses an optimized transformer architecture. Muse Glimmer is a dense causal transformer with a ded