Meta Releases Muse Glimmer, 30B Open-Weight AI Model for Local Agentic Workflows

Meta released Muse Glimmer on Monday, a 30-billion-parameter open-weight model designed for local, agentic workflows that can run on a Mac or PC with a single consumer graphics card.

“We've got even bigger models that are coming soon”

Reuters

The model is published under the Apache 2.0 license on Hugging Face, and Meta says it was trained via distillation from Muse Spark, its closed “frontier” model.

AI Business

Meta also said it will release an open-weight version of Muse Spark 1.2 in the coming weeks, after introducing Muse Spark 1.1 as a closed model for advanced reasoning and complex agentic tasks.

In a 6,500-word essay titled “The Future is for Everyone,” Mark Zuckerberg argued that superintelligence should not be concentrated in a few companies or governments, and that power should be distributed widely to empower individuals and small businesses.

Reuters reported that Zuckerberg paired the release with a call for lower U.S. barriers for open-source AI to better compete with Chinese rivals, while saying “We've got even bigger models that are coming soon.”

Zuckerberg’s essay and the Glimmer rollout framed open-weight releases as a way to keep AI from being locked behind closed systems, with Reuters quoting him saying, "Foreign labs currently hold several advantages here since American labs have to comply with many additional restrictions on training data."

In the same Reuters account, Zuckerberg argued that "U.S. policy must reduce this additional friction if we want American open source models to lead over time," and said restricting access to foreign open-source models was not an effective solution.

areacucuta

Sid Nag, president and chief research officer at Tekonyx, told AI Business that Meta is not simply switching back to open models, saying it is moving toward a hybrid approach where closed models are competitive with frontier labs while open models maximize reach to enterprises.

Arun Chandrasekaran of Gartner described Glimmer as “a very lean, lightweight model that primarily runs on your desktop,” and said Meta is “going directly toward the edge of the endpoint devices where the AI is running.”

The AI Business interview also said enterprises want control over data and infrastructure, and Chandrasekaran added that Meta “needs to build that platform layer on top of the model because that’s what enterprises want.”

Meta’s open-weight push is paired with a broader infrastructure and policy agenda, including a new $1 billion fund described by Reuters as aimed at easing fears about the impact of its data-center build-out by offering support to communities.

“Meta also plans to release the weights of Muse Spark 1.2”

Reuters

Reuters also reported that Zuckerberg said the U.S. needs to rethink policies if domestic firms are to lead in open-weight models, including on data use and distillation, and said Meta would implement a governance structure to give independent directors power to approve safety criteria for releasing models.

In a separate account, Pulso Diario San Luis said Zuckerberg warned that if “la ‘superinteligencia’ de la IA termina bajo el control de un grupo reducido de individuos, empresas, gobiernos o incluso ‘la propia IA,” then it would produce less favorable results for everyone else.

The New York Times reported that Muse Glimmer is an “open weights” model that can be freely downloaded and modified, while Muse Spark remains closed, and said Zuckerberg criticized concentrating AI power in the hands of just a few companies.

Meanwhile, the Register described Muse Glimmer as Meta’s first open-weights model in more than a year and said an open version of Muse Spark would follow, tying the release to a renewed “open weights” strategy after the earlier Llama drama.