Meta says Muse Glimmer 30B runs advanced AI on your device

Meta announced Muse Glimmer 30B, a new AI model that the company says can handle advanced, agent-like tasks directly on user devices. According to Meta, this model is designed to automate planning and workflows without sending user data to the cloud.

Why it matters

Most AI assistants today process requests on remote servers, which means user data leaves the device and companies pay for cloud infrastructure. Meta says Muse Glimmer 30B can run complex tasks, including multi-step reasoning and app integration, entirely on-device. If the model works as described, businesses could see lower infrastructure costs and fewer privacy issues.

Keeping data on the device could also help companies meet legal and customer demands for local data handling. Regulatory pressure is increasing, and data practices can affect market access. For example, the US ban of DeepSeek shows how governments are scrutinizing AI vendors over data flows.

Muse Glimmer 30B: Meta's claims and current details

Meta describes Muse Glimmer 30B as an "agentic" model, built to follow multi-step instructions, automate workflows, and interact with other apps. The company says all of this happens on the device itself, not on remote servers. Meta also claims the model works with a range of hardware, but has not released technical details or independent benchmarks. The company has not named any commercial products that will use the model first.

Other vendors, such as Nvidia, have started offering open models for local deployment. Meta is focusing on assistant-like features, aiming to support planning and automation beyond text or image generation. The real impact will depend on third-party testing and how well the model runs on actual devices. Until then, Meta's claims remain unproven outside its own labs.

What this means if you run a business

For companies using AI in customer service, internal tools, or product features, on-device models could cut infrastructure costs and speed up response times. Keeping data on the device may also make it easier to comply with data residency rules, which matter in regions with strict privacy laws or in sectors like healthcare and finance.

The effect on your business will depend on how Muse Glimmer 30B performs in real-world use and which devices support it. Many companies will wait for independent testing and integration into commercial platforms before making changes. Meta's announcement signals that device-based AI is targeting more complex tasks. Businesses should track how quickly models like this appear in the tools they use or build. If Meta's claims hold up, AI assistants could become faster, more private, and less expensive to operate than current cloud-based options.