Meta publishes the Muse Glimmer 30B technical report
Meta's research publications page posted the technical report for Muse Glimmer 30B, an on-device agentic model, with the arXiv record submitted on October 17. The paper describes a 30B-parameter causal language model with a perception encoder, quantized to about 4-bit for single-GPU inference and released under the Apache 2.0 license. Meta claims state-of-the-art results in its size class against Gemma4-31B and Qwen3.6-27B on tool use, multi-step planning and agentic task completion, though those comparisons are Meta's own.