Introducing Muse Glimmer
Meta releases Muse Glimmer, a 30B open-weights agentic model under Apache 2.0 license
“Muse Glimmer achieves strong success rates on full-task benchmarks including DeepSearch QA, MCP-Atlas, 𝛕-Bench and SWE-Bench, which measure its ability to work within scaffolds, write and debug code, and resolve multi-turn requests from start to finish.”
Meta has released Muse Glimmer, a 30B multimodal model optimized for agentic task completion, reliable tool use, and multi-step reasoning, distributed under a clean Apache 2.0 license — a notable upgrade from previous restrictive Llama licenses. The model targets local deployment on machines with 32GB+ RAM and has been validated against major agentic benchmarks including SWE-Bench and MCP-Atlas. This marks a significant re-entry by Meta into the competitive open-weights space with a model specifically designed for the agentic workflows that are now central to AI application development.