Meta announced the release of two additional artificial-intelligence models, extending the lineup that began with the Muse Spark system from its newly reorganized Superintelligence Labs division. The rollout follows the introduction of the Muse AI agent and reflects the company's intensified effort to regain a leading position in the competitive AI market.
According to Meta, internal testing on a range of industry benchmarks shows the new models can hold their own against top-tier systems such as OpenAI's GPT-6 Astra and Anthropic's Claude Opus 5.5 and Claude Fable 5.1. The company highlighted comparable performance metrics, though the specific test results were not disclosed publicly.
Meta did not provide a direct comparison with Google's Gemini 4 Argon, noting that the capabilities of that model are still being evaluated by the broader community. As a result, the relative standing of Meta's latest offerings against Gemini remains uncertain pending further external analysis.
The two models are slated to become the computational core of Meta's Muse AI agent, which has already been expanded with features like the Muse Image visual model. By integrating the new models, Muse is expected to deliver more sophisticated conversational and multimodal interactions to users across Meta's platforms.
The announcements are part of a broader strategy in which Meta has allocated billions of dollars toward talent acquisition, compute infrastructure, and research initiatives. This financial commitment underpins an expanding portfolio that aims to deliver practical AI agents to a wide audience, reinforcing Meta's ambition to be a dominant force in the AI ecosystem.