Meta Releases Muse Glimmer 30B Open Weights Coding Model
August 10, 2026
Meta Superintelligence Labs released Muse Glimmer, a 30B parameter model under Apache 2.0 optimized for local agentic workflows and function calling. It is designed to run on consumer-grade GPUs for offline tasks like local coding and LLM-as-a-judge evaluation.
HOW THIS AFFECTS YOU
●
builderYou can deploy high-performance agentic workflows on local hardware without relying on cloud APIs.
●
founderYou can build privacy-centric or offline-first AI products using a permissive open-weights model.