Meta Releases Muse Glimmer and Pushes for Open-Source AI
Meta Platforms released the Muse Glimmer AI model and a $1 billion community fund while urging the U.S. government to reduce AI regulatory barriers.
Meta Platforms released Muse Glimmer, a 30-billion-parameter open-weight AI model designed to run on personal computers with a single graphics card. The model focuses on agentic tasks such as coding, file organization, and schedule management. Alongside this launch, the company announced plans to release the weights for Muse Spark 1.2, its most advanced model, and teased a future model internally named Watermelon.
Chief Executive Mark Zuckerberg accompanied the release with a manifesto titled "The Future is for Everyone," arguing that distributing superintelligence to individuals prevents the concentration of power within a few corporations or governments. Zuckerberg urged the United States government to reduce restrictions on training data and accelerate infrastructure development to maintain a competitive lead over Chinese rivals. To manage safety, Meta is implementing a governance structure where independent directors approve safety criteria before model releases.
To mitigate local opposition to the resource-intensive nature of hyperscale computing, Meta established a $1 billion "Future is for Everyone Fund" for communities hosting its data centers. This initiative follows a model in Richland Parish, Louisiana, where tax revenues funded teacher bonuses. Meta projects spending $145 billion on AI infrastructure this fiscal year, with plans to reach $600 billion by 2028.
This strategic shift intensifies a divide in Silicon Valley. While Meta and Nvidia support open-source availability to foster innovation, rivals OpenAI and Anthropic advocate for tight restrictions to ensure safety, warning that open-weight models could be misused for cyber or biological attacks.