Meta unveils 2 new AI models for Muse Code subs and API users

Meta has released two new artificial intelligence models, Muse Wonder and Muse Wonder Full, marking a significant step in its ongoing effort to compete in the AI market. According to The Verge, these models are designed to challenge leading competitors and are currently accessible to Muse Code subscribers. The launch follows the earlier introduction of Muse Spark, the first model from Meta’s revamped Superintelligence Labs division, and the popular AI agent Muse.

How it unfolded

• Meta previously launched Muse Spark, the inaugural model from its Superintelligence Labs division.
• The company introduced the Muse AI agent, generating significant industry buzz and user interest.
• Meta released Muse Wonder and Muse Wonder Full, claiming performance comparable to industry leaders.

Who is affected

The new models are immediately available to Muse Code subscribers, allowing them to access the latest capabilities without delay. Additionally, developers and organizations can interact with Muse Wonder through the Meta Model API, providing a direct integration path for external applications. Meta states that these models can compete with leading offerings from OpenAI and Anthropic, specifically naming GPT-6 Astra, Claude Opus 5.5, and Claude Fable 5.1 as benchmarks for comparison.

According to The Verge, Meta’s claims are based on internal testing results across several industry-standard benchmarks. The company has not provided independent verification or third-party audit results for these performance metrics. It remains unclear how the new models compare to Google’s Gemini 4 Argon, as the capabilities of that model are still being defined in the market. This lack of comparative data means users must rely on Meta’s own assertions regarding competitive positioning.

What happens next

Meta intends to eventually power its Muse AI agent with the new Wonder models, beginning with subscribers before broader rollout. The company plans to develop specialized models that it can offer to everyone using the Muse platform, though no specific timeline or feature list has been shared. This strategy is part of Meta’s broader push to reestablish itself as a major force in AI, supported by billions of dollars in talent and infrastructure investments.

The firm has been steadily expanding its AI lineup, including the release of the image generation model Muse Image. Meta is also focusing heavily on bringing useful AI agents to the masses through the Muse ecosystem. These moves signal a continued aggressive approach to AI development and deployment, aiming to capture market share in both consumer and enterprise segments.

What to do and how to stay safe: Meta

• Review current AI model usage in your organization to assess compatibility with new Meta offerings.
• Monitor the Meta Model API documentation for updates on Muse Wonder integration capabilities.
• Evaluate internal benchmark results against Meta’s claims before adopting the new models for critical tasks.
• Stay informed about Meta’s timeline for rolling out Muse Wonder to the broader Muse agent user base.

Step-by-step guide: Injection attacks examples: how prompt leaks break AI systems

General security guidance from the Circuit Observer newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

What are the new AI models launched by Meta?

Meta has launched Muse Wonder and Muse Wonder Full, which are available for Muse Code subscribers and via the Meta Model API.

Which existing models does Meta claim its new models rival?

Meta claims its new models can compete with OpenAI’s GPT-6 Astra, Anthropic’s Claude Opus 5.5, and Claude Fable 5.1.

Will the Muse AI agent use these new models?

Yes, Meta plans to power its Muse AI agent with the Wonder models, starting with subscribers, though no timeline is provided.