Meta launches open-weight Muse Glimmer, Mark Zuckerberg says he wants superintelligence for all

Meta has released Muse Glimmer, a 30-billion-parameter open-weight AI model designed to run locally on consumer hardware. The model is built for AI agents that can use tools, reason through tasks and work with text and images without relying entirely on the cloud.Meta releases 30B Muse Glimmer for local AI. (Photo: Reuters)Meta is making another big bet on open AI — and this time, it wants powerful AI agents to run directly on your computer. The company has released the weights of Muse Glimmer, a 30-billion-parameter AI model designed to run locally on consumer hardware, allowing developers to build AI agents without relying entirely on cloud servers.Mark Zuckerberg highlighted the move in a post on X, saying Meta is “a strong supporter of open source” and that he was proud of the release. He also confirmed that Meta plans to release the weights of Muse Spark 1.2, its latest foundation model, soon.A 30B model that can run locallyOne of the biggest highlights of Muse Glimmer is its size; it is a 30-billion-parameter model that is designed for local AI agents. In simpler terms, developers can run it on compatible Macs and PCs instead of sending every request to a remote data centre.Meta says the model can run on a machine with a single consumer GPU. The company has also compressed the model to roughly 4-bit precision, bringing its size down to under 20GB. This makes it possible to fit the model within the memory limits of some high-end consumer hardware. In other words, AI can now live outside the cloud. A local model could be useful for tasks where privacy, offline access or control over data matters. Developers could also build assistants that work with files, coding tools and other applications while keeping the AI processing closer to the user.This isn't just another chatbotMuse Glimmer is designed to do more than answer questions. Meta calls it an agentic model, which means it is trained to work through tasks involving multiple steps. It can use tools, maintain a plan, work with images, write and debug code, and recover when something goes wrong. That could allow developers to build agents that do things rather than simply explain how to do them.For instance, instead of asking an AI how to organise a set of files, a developer could build an agent that can actually inspect those files, use the required tools and complete the task.Meta is betting on open AI againThe release also marks an important moment for Meta's broader AI strategy.The company has built much of its AI identity around open-weight models, particularly through its Llama family. But Meta had previously held back the full release of its top-tier Muse Spark model amid internal safety debates. Now, Meta is opening Muse Glimmer and preparing to release Muse Spark 1.2 as well.That suggests Meta is once again leaning heavily into the idea that developers should be able to access and build on its AI models rather than being limited to closed cloud services.Zuckerberg made that position clear in his X post, saying Meta is a “strong supporter of open source” while praising the team behind the models. The model release also fits into a much bigger argument Zuckerberg has been making about the future of AI.In a separate essay, the Meta CEO argued that superintelligent AI should not be controlled exclusively by companies or governments. That puts Meta's approach in direct contrast with the more closed strategies used by companies such as OpenAI, Anthropic and Google.- EndsPublished On: Aug 10, 2026 17:18 IST