Muse Glimmer is a 30-billion-parameter model with open weights (Apache 2.0) built for continuous agentic workloads on local hardware. It targets systems with 24 GB VRAM and focuses on long-running agent workflows rather than one-off queries, with developers reporting strong results on agentic benchmarks in its class. GGUF quantizations are already available on Hugging Face, alongside an official guide for local deployment. Separately, Muse Spark 1.2 is in the works, with its weights promised to be opened later.

Related: Meta Releases Multimodal AI Model Muse Spark, Superintelligence Labs Releases Muse Spark 1.1, a Multimodal Agent Model

Muse Glimmer GGUF on Hugging Face · Muse Glimmer guide on unsloth.ai