Meta Launches New AI Model, Zuckerberg Urges Open-Source Push

Meta CEO Mark Zuckerberg is pushing for the United States to lower barriers for open-source artificial intelligence models, asserting that such a move is crucial for American firms to effectively compete with Chinese rivals. This call comes as Meta unveiled a new open-weight model, Muse Glimmer, and announced plans for more such releases.

The newly released Muse Glimmer is notably smaller than leading AI models from competitors. Its design focuses on agentic tasks and allows it to run on standard personal computers with a single graphics card. This aims to meet the growing demand for AI systems that can operate directly on users' devices. Open-weight models are generally more affordable than those from major AI labs like OpenAI and Anthropic. They also offer publicly accessible core components, facilitating easier customization, a contrast to closed models that companies maintain under strict control.

Meta's latest move occurs as the social media giant seeks to bolster its standing in the high-stakes AI race, having established a new, costly superintelligence team last year. Zuckerberg's public stance also signifies strong support for open-weight AI, which is gaining favor among businesses concerned about escalating AI costs and recent cybersecurity incidents involving models from various leading companies. Hugging Face, an AI coding platform, recently reported using a Chinese open-weight model to defend against a cyberattack, noting that closed-source models have limitations for cybersecurity work.

Zuckerberg emphasized that U.S. policy requires rethinking if domestic companies are to lead in the open-weight AI sector. He pointed out that Chinese startups are currently ahead in this area, with models from Moonshot, Alibaba, and DeepSeek demonstrating performance comparable to top U.S. AI labs. He stated that foreign labs possess advantages partly because American labs face more restrictions on training data. "US policy must reduce this additional friction if we want American open source models to lead over time," Zuckerberg said, adding that restricting access to foreign open-source models is not an effective strategy. Meta's stock saw a slight increase in premarket trading following the announcement.

“

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.