Meta launched its newest family of AI models, Muse Glimmer—designed to run locally on laptops—on Monday, alongside a 6,500-word essay from Mark Zuckerberg championing the cause of American open-source AI.

Geopolitical Stakes and Rivalry

In the essay, Zuckerberg argues that powerful AI should not be controlled by a handful of companies, a clear jab at closed-source competitors like OpenAI and Anthropic. He positions Meta’s commitment to open-source as critical to challenging Chinese models from labs like DeepSeek and Moonshot, which have been closing the gap with American technology.

Zuckerberg also urged Washington to support domestic efforts, stating that restricting access to foreign open-source models is not an effective solution. Instead, he argued that the goal should be for American open-source models to lead globally by removing hurdles that hinder their competitiveness.

Opening the Weights of Muse Spark 1.2

As part of this renewed commitment, Meta announced plans to open the weights for Muse Spark 1.2, its most advanced AI model currently on the market. By releasing the underlying trained parameters, Meta will allow developers to download, inspect, and modify the model, rather than accessing it solely through a proprietary API.

The company revealed that the Glimmer series was trained using distillation, a process where a smaller model learns from the larger proprietary Muse Spark system. This development comes as Meta heavily funds its new Superintelligence Labs, which recently recruited former Scale AI CEO Alexander Wang.

A Strategic Pivot

The refocus on open-source reflects Meta's attempt to regain momentum after its recent models fell short of systems from OpenAI and Anthropic. By leaning into this strategy, Meta aims to:

  • Differentiate itself from the closed-door approach of its primary rivals.
  • Exert influence over the developer ecosystem even as others capture proprietary revenue.
  • Reassure investors of its continued competitiveness in the frontier lab space.
  • Build a political case for government and user support through transparency.