Meta releases its first frontier model, Muse Wonder
Meta Superintelligence Lab releases Muse Wonder and Muse Wonder Full, its first frontier AI models featuring advanced coding and memory capabilities.
Stock photo for illustration only, not from the actual event
- Meta Superintelligence Lab introduces Muse Wonder and Muse Wonder Full, its first frontier AI models.
- Achieves benchmark scores comparable to Fable 5.1 and GPT-6 Astra with strong coding performance.
- Muse Wonder Full is restricted to Muse Code subscribers and unavailable via API.
- Trained at Meta's Prometheus data center located in central Ohio, United States.
Meta has taken a major step in artificial intelligence by unveiling its first-ever frontier AI models through its Superintelligence Lab, featuring two variants named Muse Wonder and Muse Wonder Full. This release marks a strategic shift for the company, following its previous rollout of the lightweight Muse Spark model which focused primarily on low latency and quick response times.
The newly introduced Wonder models are larger in scale and built to handle significantly more complex computational tasks. According to internal benchmark testing conducted by Meta, both Wonder and Wonder Full achieve performance scores comparable to industry competitors Fable 5.1 and GPT-6 Astra, showing particularly strong results in complex domains such as software programming and code generation.
In terms of commercial accessibility, accessing Muse Wonder through the Meta Model API will cost 4 dollars per million input tokens and 20 dollars per million output tokens, positioning it at a price point roughly equivalent to GPT-6 Sol and Opus 5.5. Meanwhile, the larger Muse Wonder Full model will be restricted exclusively to Muse Code subscribers and will not be served through the API. Both models will also serve as the underlying technology powering the Muse agent.
Stock photo for illustration only, not from the actual event
To demonstrate improvements in recall capabilities, Meta highlighted results from the MemoryBench benchmark, which evaluates continual learning and long-term memory construction frameworks. Company testing indicates that Muse Wonder achieves higher scores on MemoryBench than competing models including Fable, Opus, and Astra.
"We put a tremendous amount of effort into making sure your model can remember what’s important to you over days, weeks, months, and beyond."
Kavya Kopparapu, Research Scientist at Meta
This rollout represents Meta's first state-of-the-art model release since the restructuring of its artificial intelligence division in August of last year under Scale AI founder Alexandr Wang. Since that reorganization, the company had concentrated heavily on lightweight models and agent systems rather than competing directly in the heavy-duty frontier model space.
The introduction of Muse Wonder highlights Meta's ambition to compete directly with leading AI developers in the heavy-duty model category. Utilizing the company's newly operational Prometheus data center in Ohio underscores the massive infrastructure investments required to train and deploy state-of-the-art frontier architectures at scale.
According to the company announcement, the new models were trained at Meta's Prometheus data center situated in central Ohio, United States, which officially came online earlier this year.
Source: TechCrunch
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment