Meta has officially entered the competitive landscape of 'frontier' AI models with the release of Muse Wonder and Muse Wonder Full. These models, developed by the company’s Superintelligence Labs, are positioned as high-performance tools capable of rivaling the industry's leading AI systems. According to company announcements, the models are available immediately to Muse Code subscribers, while the standard Wonder model is also accessible via the Meta Model API.

There is some variation in how these models are characterized. TechCrunch identifies this release as Meta’s first official attempt at a 'frontier model,' a term used to describe the most advanced and capable AI systems currently in development. Wired emphasizes the safety features included in the release, noting that the models contain specific mechanisms designed to prevent AI agents from behaving unpredictably, often referred to as 'going off the rails.'

Meta has outlined a roadmap for the integration of these models into its broader ecosystem. The company intends to eventually power its existing Muse AI agent with the Wonder technology, beginning with its subscriber base. Future plans include the development of specialized versions of these models intended for a wider public release, though specific timelines for these broader deployments have not been finalized.