Introducing Muse Wonder and Muse Wonder Full: Frontier Models for Agents
Meta unveils Muse Wonder and Muse Wonder Full, its most advanced AI models designed for agentic tasks across domains like law, finance, and health, featuring Active Memory and Active Focus capabilities.
Meta introduces the Muse Wonder series, comprising Muse Wonder and Muse Wonder Full, as its most capable AI models to date, built on the foundation of Muse Spark. These models are engineered to handle complex agentic tasks, such as processing insurance claims, interpreting leases, or analyzing financial documents, by interpreting information and solving problems autonomously.
The new models incorporate Active Memory, which retains relevant user-specific details across conversations, and Active Focus, enabling them to persist through challenging tasks, adapt approaches, recover from errors, and seek clarification when necessary. This ensures they can follow through on goals while maintaining accuracy and reliability in real-world applications.
Muse Wonder and Muse Wonder Full are designed to operate securely, refusing requests for harmful actions and preventing unauthorized sharing of private information. They can assist developers and security teams in identifying code vulnerabilities and strengthening defenses against cyber threats without compromising ethical boundaries.
Meta states that Muse Wonder is available immediately via the Meta Model API and Muse Code, with broader integration into the Muse platform planned for subscribers. The company adheres to its Meta Superintelligence Scaling Framework for training, evaluating, and releasing these models, with a detailed safety report forthcoming.