Meta has launched a new AI model called “Muse Glimmer” as part of its efforts to expand open-weight artificial intelligence.
Meta CEO Mark Zuckerberg has urged the U.S. government to reduce regulatory barriers for open-source and open-weight AI models so American companies can better compete with Chinese rivals.
Open-weight AI models are generally cheaper than advanced models developed by leading AI companies such as OpenAI and Anthropic. They also provide access to key components that allow users and businesses to customize them, unlike closed models that remain fully controlled by their developers.
The new Muse Glimmer model is much smaller than many leading AI models from competing companies. It is designed for agentic workflows and can run locally on a Mac or PC using a single consumer graphics card with about 24 GB of VRAM.
Zuckerberg said the U.S. should reduce regulatory restrictions on domestic open-source and open-weight AI developers. He warned that excessive rules could give Chinese companies an advantage as they continue to make progress in open AI development.
Referring to open-source models, Zuckerberg said foreign AI labs currently have some advantages because U.S. companies face additional restrictions on training data. He said U.S. policies should reduce these challenges if American open-source models are to remain competitive.
Zuckerberg also said limiting access to foreign open-source models would not be an effective solution.
His comments come as support for open-weight AI continues to grow. Businesses are becoming more concerned about rising AI costs as well as recent cybersecurity incidents involving models from companies including Anthropic, OpenAI and Meta.
Meta’s latest launch comes as the company works to strengthen its position in the AI industry after creating an expensive new superintelligence team last year to compete in the growing AI race.
Along with the open-weight model, Meta introduced a new safety governance system that includes an independent oversight board. The board will approve safety standards for future open models as the company seeks to balance open access with safety concerns.
Zuckerberg said making AI models more accessible can help improve security, customization and innovation.
Chinese startups are also making strong progress in open-weight AI models. Moonshot’s Kimi K3, Alibaba’s Qwen3.8-Max and DeepSeek’s V4-Flash are reportedly delivering performance comparable to leading systems from U.S. AI companies. In contrast, major models from OpenAI, Anthropic and Alphabet’s Google remain closed source.
In addition to launching Muse Glimmer, Meta announced plans to release open-weight versions of its flagship Muse Spark model.
