Meta launches new AI model as Zuckerberg champions open-weight push

Meta Platforms CEO Mark Zuckerberg on Monday called for the United States to reduce regulatory barriers around open-weight artificial intelligence, arguing that a more flexible approach would help American companies compete with rapidly advancing Chinese AI developers.

His comments came as Meta unveiled a new open-weight model, Muse Glimmer, and indicated that larger models are also in development.

Muse Glimmer is smaller than many of the leading AI systems developed by competitors and is designed to perform agentic tasks on personal computers, including Macs, using a single graphics card.

Zuckerberg said the company plans to introduce more powerful models soon, while arguing in a 14-page essay titled “The Future is for Everyone” that AI should be broadly accessible rather than concentrated in the hands of a small number of companies or governments.

Open-weight AI models have attracted growing interest from businesses seeking lower-cost and more customizable alternatives to closed systems operated by companies such as OpenAI and Anthropic.

Unlike closed models, open-weight systems make key components available for modification and deployment, although concerns over cybersecurity and misuse remain part of the wider debate surrounding the technology.

Zuckerberg also pointed to the growing strength of Chinese companies in the open-weight AI sector, citing models developed by firms including Moonshot, Alibaba and DeepSeek.

He argued that US developers face additional restrictions involving training data and other areas, giving foreign competitors an advantage. According to Zuckerberg, American policy needs to reduce such obstacles if US companies are to remain leaders in open AI.

Meta is also preparing to release the weights of Muse Spark 1.2, described as its most advanced model and developed by a costly superintelligence team established last year.

At the same time, Zuckerberg announced a $1 billion fund aimed at supporting communities affected by the expansion of data centers, as opposition to large AI infrastructure projects continues to grow in parts of the United States.

Zuckerberg said the US also needs to make it easier to build the infrastructure required to support AI development, noting that Meta could spend as much as $145 billion on AI infrastructure this year. He further said Meta would introduce a governance system giving independent directors authority over safety standards for model releases.

His comments come as the Trump administration considers its approach to open-weight AI, with officials recently indicating that such models would not be subjected to voluntary safety testing.