Meta Debuts Muse Wonder Frontier AI Model
Meta’s Superintelligence Lab has released Muse Wonder and Muse Wonder Full, representing the company’s first entry into frontier artificial intelligence models, according to TechCrunch. The larger models follow the earlier release of the lightweight Muse Spark and feature improved capabilities for complex tasks such as coding.
TechCrunch reports that the models were trained at Meta’s Prometheus data center in central Ohio, which came online earlier this year. According to company testing cited in the report, the models achieve benchmark scores comparable to competing systems like Fable 5.1 and GPT-6 Astra.
The standard Muse Wonder model is accessible through the Meta Model API at a cost of $4 per million input tokens and $20 per million output tokens. The larger Muse Wonder Full variant is restricted to Muse Code subscribers and will not be served through the API. Both models power the Muse agent and emphasize active memory features designed to retain and synthesize user information over extended periods.
The release marks Meta’s first frontier model deployment since the restructuring of its artificial intelligence division under Scale AI founder Alexandr Wang in August of last year. Both models were designed to support personal assistant applications with advanced continual learning capabilities.
Based on reporting by techcrunch.com.
