Meta Releases Muse Glimmer, a 30B Open-Weight Model Under Apache 2.0

Meta has introduced Muse Glimmer, a new 30B parameter language model released under the permissive Apache 2.0 license, marking a departure from the company's previous Llama licenses, which were criticized for their restrictions. The model is optimized for local use, as demonstrated by Simon Willison, who generated a pelican image using LM Studio's 18.16 GB version of the model and tested it with his llm-coding-agent plugin against a fresh checkout of Datasette. The plugin made multiple tool calls to explore the codebase, and the response was captured in a long transcript. Willison, who runs a blog on LLM developments, praised the model's size, noting that it leaves ample RAM for other applications on machines with 32 GB or more (he has 128 GB). The post, published on 10th August 2026, is a link post and includes a call for sponsorship at $10 per month for a curated email digest of monthly LLM developments. This release signals Meta's renewed commitment to open-weight models, potentially impacting the AI community's reliance on proprietary systems.