Meta released a new open-weight AI model on Monday and CEO Mark Zuckerberg used the launch to call for lower US barriers to open-source AI, arguing American labs are handicapped in the race against Chinese rivals.
The model, Muse Glimmer, is much smaller than frontier systems from OpenAI and Anthropic. It is designed for agentic tasks and can run on a Mac or PC with a single graphics card, tapping demand for AI that runs directly on people's devices. The New York Times reports it is nearly identical to Muse Spark, Meta's most powerful model, released as an open-weight version — a step Zuckerberg has championed as he pushes Meta back into the frontier race after forming a costly superintelligence team last year.
Open-weight models are typically cheaper than closed frontier models and come with publicly accessible core components for easy customization. Zuckerberg argued the US needed a policy rethink: 'Foreign labs currently hold several advantages here since American labs have to comply with many additional restrictions on training data. US policy must reduce this additional friction if we want American open source models to lead over time,' he said, adding that restricting access to foreign open-weight models was not an effective solution.
Chinese startups currently lead the open-weight race — Moonshot's Kimi K3, Alibaba's Qwen3.8-Max and DeepSeek's V4-Flash deliver performance rivaling top US systems, while the leading models of OpenAI, Anthropic and Google are closed source. The Trump administration told AI developers earlier this month that it will not put open-weight models through voluntary safety tests, according to two sources.
Zuckerberg also advocated model distillation — using a powerful system to train a smaller one — and said Meta will adopt a governance structure giving independent directors power to approve safety criteria for releasing models. Meta shares rose about 1% in premarket trading.
