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

Muse Glimmer is a 30B-parameter dense model released under Apache 2.0, built for always-on local agent workflows, with a 2B vision encoder, 28B text decoder, and 120K+ context. Hugging Face shipped day-0 support in transformers, llama.cpp, and vLLM; NVIDIA reports 20K tokens/sec on a single GPU.