Reproduce “Muse Glimmer 30B: An On-Device Agentic Model” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Arash Ebrahim Abtahizadeh, Muna Aghamelu, Adil Ahmed, Juan C. Angeles Acuna, Alexis Audran-Reiss, Asaf Avrahamy, Yoram Bachrach, Can Balioglu and 321 more
Muse Glimmer distills Muse Spark into a dense 30B model for agentic workflows, then compresses it for local deployment. Quantization and speculative decoding help it fit consumer hardware. MCPAtlas tests multi-turn tool use across 500 tasks. Muse Glimmer scores 75.5, versus 54.2 and 62.5 for similarly sized thinking-mode models; results vary across evaluations and do not establish success on arbitrary real-world tasks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.