Muse Glimmer: Meta's new open-weight AI model runs on PCs
What's the story
Meta has launched its latest artificial intelligence (AI) model, Muse Glimmer. With 30 billion parameters, the new addition is much smaller than other leading AI models in the market. It is specifically designed for agentic tasks and can run on a Mac or PC with just a single graphics card. This move comes as part of Meta's strategy to tap into the growing demand for AI systems that can run directly on people's devices.
Performance
Meta has released weights on Hugging Face
Meta has released open weights and documentation for Muse Glimmer on Hugging Face to streamline agent development.
Outperforming Gemma4-31B and Qwen3.6-27B across top benchmarks like MCP Atlas and SWE-Bench Pro, the model features an integrated DFlash-based "drafter" network.
This companion model proposes token blocks for parallel verification by the main engine, accelerating generation speeds over traditional token-by-token processing while maintaining identical output quality.
Policy push
Zuckerberg calls for policy changes to support open-weight models
Meta CEO Mark Zuckerberg has called for a rethink of US policies to allow domestic firms to take the lead in open-weight models.
He said, "Foreign labs currently hold several advantages here since American labs have to comply with many additional restrictions on training data."
The statement comes as part of his support for open-weight AI, which is gaining traction amid rising cybersecurity concerns and high costs associated with proprietary models.
Economic advantage
Economic advantages of open-weight models
Open-weight models are generally cheaper than those from frontier labs like OpenAI and Anthropic.
They also provide publicly available core components for easy customization, unlike closed models that companies keep fully under their control.
The economic edge of open-weight models is another reason why Zuckerberg is pushing for policy changes to promote their development in the US.
Safety measures
AI model distillation and governance structure
Zuckerberg also emphasized the importance of AI model distillation, or using a powerful AI system to train a smaller one.
He said that Meta would implement a governance structure to give its independent directors the power to approve the safety criteria for releasing models.
This move is part of Meta's broader strategy to ensure the safe and responsible development and deployment of its AI technologies.