Meta launches Muse Wonder models for agentic tasks
Meta's latest AI models are built to handle complex, multi-step work across messages, calendars and documents. Designed to remember context and keep pushing when tasks get hard, they arrive first to subscribers, while the safety story is still unfolding.
Meta launches Muse Wonder models for agentic tasks
Meta is introducing the Muse Wonder series, Muse Wonder and Muse Wonder Full, which it calls its most capable models to date. Meta says they deliver state-of-the-art performance on personal and professional agentic tasks, building on the Muse Spark groundwork laid over the past fourteen months. For users, that means an agent that can handle multi-step work across messages, calendars, and documents.
Key Features at a Glance
Active Memory: Carries what matters across conversations and brings back relevant details for the task at hand, useful when work spans multiple sessions.
Active Focus: Keeps working toward a goal when a task gets hard, tries other approaches, recovers from errors, and asks for clarification before following through on anything important.
Expert domains: Interprets information in law, finance, and health, such as an insurance claim, a lease, or someone's finances.
Frontier coding: Performs well for people building everything from a game to an app for a new business.
Privacy safeguards: Trained to avoid sharing private information without permission and to turn down requests to assist with actually harmful acts.
Security work: Developers and security teams can still use the models to find flaws in their own code and strengthen their systems.
Availability
Muse Wonder is available today through the Meta Model API and Muse Code, and it will soon power more of Muse, rolling out first to subscribers; a full report with additional safety evaluations is due soon.
