Meta has released Muse Glimmer, a smaller open-weight AI model aimed at agentic tasks.
Zuckerberg wants the US to reduce barriers for American companies developing open-weight models.
Meta is looking to strengthen its position in the AI race as Chinese models gain ground.
Meta CEO Mark Zuckerberg has called for lower US barriers for open-weight artificial intelligence (AI) models, arguing that American companies need a more level playing field to compete with Chinese rivals. His comments came as Meta released a new AI model, Muse Glimmer, and said it plans to launch more open-weight models.
Unlike large frontier models that require significant computing power, Muse Glimmer is designed for agentic tasks and can run on a Mac or PC with a single graphics card, according to a Reuters report. Meta is targeting growing demand for AI systems that can operate directly on users’ devices.
1 August 2026
Get the latest issue of Outlook Business
Open-weight models generally allow developers to access core components and customise them, while closed models keep those components under the developer’s control. They can also be cheaper to operate than leading models from companies such as OpenAI and Anthropic.
Zuckerberg Pushes For Policy Change
Zuckerberg said US policy needs to reduce what he described as additional restrictions faced by American companies training open-weight models. He argued that foreign AI labs currently have advantages because US companies face greater restrictions on training data.
The Meta CEO also said restricting access to foreign open-weight models would not be an effective solution. Instead, he called for policies that would help American open-source AI models compete over the long term.
Advertisement
His comments come as Chinese companies gain ground in open-weight AI. Reuters reported that models from Moonshot, Alibaba and DeepSeek are delivering performance that rivals some leading systems from US AI developers.
Meta Steps Up AI Push
Meta has been investing heavily in AI as it seeks to strengthen its position in the sector. The company formed a new superintelligence team last year as part of that effort.
Zuckerberg also backed AI model distillation, a process in which a powerful AI model is used to train a smaller one. Meta plans to establish a governance structure under which independent directors can approve safety criteria for releasing models, he said.
Advertisement
The push comes as businesses become more concerned about rising AI costs and cybersecurity risks involving AI systems.
