Meta Releases New Open-Weight AI Model, Zuckerberg Pushes for Policy Change

Meta launched a new open weight AI model called Muse Glimmer on Monday, notably smaller than leading systems from rivals and designed specifically for agentic tasks that can run on a single graphics card in a standard Mac or PC. The move signals Meta’s push to compete in the growing demand for AI tools that operate directly on personal devices rather than relying heavily on cloud infrastructure.

Zuckerberg Calls for Policy Reform

Meta CEO Mark Zuckerberg argued that the United States needs to ease restrictions on open-source AI development to help domestic companies compete with Chinese rivals, who currently lead the open-weight model race with systems from firms like Moonshot, Alibaba, and DeepSeek. He said American labs face additional training data restrictions that put them at a disadvantage, while cautioning that blocking access to foreign open-source models isn’t an effective countermeasure.

Industry Momentum Behind Open-Weight Systems

Open weight models, generally cheaper and more customizable than closed alternatives from major AI labs, have gained traction as businesses grow cautious about rising AI costs and recent security incidents involving AI systems. Meta also announced plans for a governance structure allowing independent directors to approve safety criteria before releasing future models. The launch follows the company’s formation of a new superintelligence team last year aimed at strengthening its position in the AI race.