Meta launches Muse Glimmer as Zuckerberg calls for fewer US AI curbs to beat China
Meta has launched Muse Glimmer, a new open-weight artificial intelligence model designed to run on a single graphics card, as chief executive Mark Zuckerberg called for fewer US restrictions on open AI technology to help American companies compete with Chinese rivals.
The smaller model is designed to perform agentic tasks directly on Macs and PCs, rather than relying entirely on cloud-based computing. The launch comes as businesses increasingly seek cheaper AI systems that can run on their own devices.
“We’ve got even bigger models that are coming soon,” Zuckerberg said in a video post accompanying a 14-page essay titled “The Future is for Everyone”.
“The notion AI is so dangerous that the only safe path is an extreme concentration of power seems inherently problematic.”
What is Meta’s Muse Glimmer AI model?
Muse Glimmer is an open-weight model that allows developers greater access to its core components, making it easier to customise and deploy for specific applications. Meta said the model is built to run on a single graphics card and is focused on agentic AI tasks.
The approach differs from closed models offered by companies such as OpenAI and Anthropic, where the underlying components remain under the developers’ control.
Open-weight models have gained traction as companies look for ways to reduce rising AI costs. They have also become increasingly important in the US-China competition over AI, with Chinese developers producing powerful and relatively inexpensive models.
Zuckerberg calls for changes to US AI policy
Zuckerberg used the launch to argue that US policy needs to change if American companies are to remain competitive in open-weight AI.
“Foreign labs currently hold several advantages here since American labs have to comply with many additional restrictions on training data,” Zuckerberg said, referring to open-source models.
“U.S. 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 was not an effective solution.
He also defended AI distillation, a technique in which the outputs of a powerful AI model are used to train a smaller model that can perform some of the same tasks with less computing power.
Meta plans to establish a governance structure that would give independent directors the authority to approve safety criteria for releasing its AI models.
Meta backs open AI as Chinese rivals gain ground
Mark Zuckerberg’s latest comments come amid an intensifying debate over whether advanced AI should remain tightly controlled or be made more widely available.
“Rather than centralizing superintelligence, we should distribute it,” Zuckerberg wrote in his essay. He added, “This has the potential to begin a new era of personal empowerment.”
Chinese companies have made significant advances in open-weight AI, with models from developers including Moonshot, Alibaba and DeepSeek gaining attention from businesses seeking alternatives to expensive proprietary systems.
Zuckerberg also highlighted China’s advantage in building infrastructure.
“One significant disadvantage that the U.S. has compared to countries like China is that it is more difficult to build infrastructure here,” he said.
Meta is expected to spend as much as $145 billion this year on AI infrastructure.
Meta plans more AI models
Meta has historically been one of the strongest US technology companies supporting open AI models. However, it recently shifted towards closed systems after its previous models struggled to match leading offerings from OpenAI and Anthropic.
Its latest frontier model, Muse Spark, remains closed and is being sold to users. Meta also plans to release the weights of Muse Spark 1.2, its most advanced model.
In the coming months, Meta plans to release another AI model, internally known as Watermelon, which Zuckerberg said will be more powerful than Muse Spark. It remains unclear whether that model will be open-weight or closed.
Zuckerberg also announced a $1 billion fund to support communities affected by Meta’s expanding data-centre infrastructure.
He has argued that faster infrastructure development is essential for the US to maintain its AI advantage over China, adding another dimension to Meta’s push for a more open AI ecosystem.
