Meta Launches Open-Weight Model, Urges US Policy Shift

Meta released a new open-weight AI model on 10th August and chief executive Mark Zuckerberg called for lower U.S. regulatory barriers to help American firms compete with Chinese developers that have taken the lead in open-source artificial intelligence.

The new model, Muse Glimmer, is compact enough to run agentic tasks on a personal computer with a single graphics card. Zuckerberg said larger models are coming soon and confirmed that Meta will release the weights of Muse Spark 1.2, its most advanced system, built by the superintelligence team formed last year. Meta shares rose nearly 3 percent in premarket trading.

The launch came alongside a 14-page essay titled "The Future is for Everyone," in which Zuckerberg argued that concentrating AI in a few closed systems is inherently risky. He said the recent cybersecurity breach at Hugging Face, where a rogue OpenAI model was countered using a Chinese open-weight model because closed models had cybersecurity restrictions, showed why open access matters.

Chinese firms Moonshot, Alibaba, and DeepSeek now offer open-weight models that rival top U.S. systems in performance at much lower cost. Zuckerberg said U.S. labs face restrictions on training data and distillation that foreign labs do not. "U.S. policy must reduce this additional friction if we want American open source models to lead over time," he wrote, adding that restricting access to foreign models was not an effective solution.

Zuckerberg also announced a $1 billion fund to ease community concerns about data centre expansion, an issue that has become a hurdle for AI infrastructure in the U.S. Meta plans to spend up to $145 billion on AI infrastructure this year.