Editorial illustration for Meta Unveils Muse Wonder AI Models for Code Subscribers
Meta Launches Muse Wonder AI for Code Subscribers
Meta rolled out two new AI models on Tuesday, Muse Wonder and Muse Wonder Full, available immediately to Muse Code subscribers. Wonder can also be reached through the Meta Model API, the company said. It's the latest move from Meta's Superintelligence Labs division, which already shipped Muse Spark and the AI agent Muse earlier this year as part of a broader attempt to catch up after months of watching OpenAI and Anthropic set the pace.
Meta says internal benchmark testing shows the new models holding their own against GPT-6 Astra, Claude Opus 5.5, and Claude Fable 5.1. How they compare to Google's Gemini 4 Argon is harder to say, since that model's full capabilities are still being sorted out by outside testers. Meta hasn't said when Wonder will actually power the Muse agent for subscribers, or when a version might reach the wider Muse user base, beyond saying it plans to build specialized models for that purpose eventually.
The release follows billions in spending on AI infrastructure and talent, part of a company-wide push that also includes the image model Muse Image.
Why this matters
Meta is betting that Muse Wonder and Muse Wonder Full can close the gap with whatever OpenAI, Anthropic, and Google ship next, and it's doing so by bundling the models into Muse Code rather than releasing them as standalone products. That's worth watching closely. Access through the Meta Model API means developers can test Meta's claims directly instead of relying on benchmark charts in a press release.
For founders building on Muse, the real question is whether Wonder actually powers the agent better than whatever came before it, or whether this is another incremental swap dressed up as a leap. Meta has a habit of naming things aggressively (Wonder, Wonder Full) while staying vague on the technical specifics that researchers actually care about, like training data, context limits, or where it falls short against GPT or Claude on real tasks. Until independent benchmarks or developer feedback surface, treat "toe-to-toe with the industry's best" as a marketing line, not a verdict.
Subscribers get first access today; everyone else should wait for the reviews.
Related Reading
Browse all Meta news →
Business & Startups Nordic pilot adds Gemini for Education, NotebookLM to boost AI literacy
Business & Startups Kling launches Video O1, all-in-one model with MVL bridge using transformer
Business & Startups DeepSeek Seeks More Capital Weeks After USD 7B Funding Round
Related Topic Meta’s Muse Tracks Your Email and Bank Data While You Work
AI Tools & Apps
Related Topic Meta Faces Lawsuit Over AI Training Data, Facial Recognition
Policy & Regulation
From Archives Meta Announces First Model From Its New AI Lab, Muse Spark
From Archives Microsoft's Andreou: OpenAI Models Excel in Specific Tasks
From Archives Meta Launches Its First 'Frontier' AI Model, Muse Wonder
Research & Benchmarks
From Archives Meta's AI Logo Designer Faces Backlash After Revealing Workshop Fee
AI Tools & Apps
Common Questions Answered
What are the two new AI models that Meta released for Muse Code subscribers?
Meta unveiled Muse Wonder and Muse Wonder Full, which are now available immediately to Muse Code subscribers. These models can also be accessed through the Meta Model API, providing developers with direct access to test Meta's latest AI capabilities.
How do Muse Wonder and Muse Wonder Full compare to competing models from OpenAI and Anthropic?
According to Meta's internal benchmark testing, the new Muse models can rival leading models from OpenAI and Anthropic, including GPT-6 Astra, Claude Opus 5.5, and Claude Fable 5.1. Meta points to results from several industry benchmarks to support these claims about competitive performance.
What is Meta's strategy for distributing Muse Wonder instead of releasing it as a standalone product?
Meta is bundling Muse Wonder and Muse Wonder Full into Muse Code rather than releasing them as independent products. This approach allows developers to access and test the models through the Meta Model API, enabling them to verify Meta's performance claims directly instead of relying solely on benchmark charts from press releases.
Which division at Meta is responsible for developing the new Muse Wonder models?
Meta's Superintelligence Labs division developed the Muse Wonder models as part of their broader effort to catch up in the AI race. This division has already shipped Muse Spark and the AI agent Muse earlier in the year, demonstrating Meta's commitment to competing with OpenAI and Anthropic.
Further Reading
- Meta releases its first frontier model, Muse Wonder - TechCrunch
- Meta Unveils New Series of Advanced AI Models - The New York Times
- Meta Puts Muse Back at the AI Frontier - WIRED
- Meta debuts Muse Wonder AI model in test of users' willingness to pay - CNBC
- Introducing Muse Wonder and Muse Wonder Full: Frontier Models for Agents - Meta
