Meta releases Muse Glimmer, a 30B open-weight agentic model

Muse Glimmer is a 30B-parameter dense multimodal model with a 120K+ context window, optimized for local agentic workflows and released under Apache 2.0. It runs on a single consumer GPU (24GB VRAM) and delivers 20K tokens/sec on NVIDIA hardware. Day-0 support ships in transformers, llama.cpp, vLLM, and Inference Endpoints.