Introducing Muse Glimmer: An Open Agentic Model That Runs on Your Device
Summary
Meta announced Muse Glimmer, a 30B open-weight local agentic model designed for on-device operation. It runs on consumer hardware with quantization to 4-bit to fit under ~20 GB and uses a DFlash drafter to speed up generation. The model is open-sourced on Hugging Face with documentation and integrations planned for llama.cpp, MLX, and ExecuTorch, enabling offline or edge deployments for local agents.