Meta Launches Muse Wonder and Muse Wonder Full for Agentic Work

Meta has released Muse Wonder and Muse Wonder Full for personal and professional agentic tasks, with API access today and a subscriber rollout planned. Creators and operators should assess privacy, coding, and automation workflows now.

What happened

Meta is introducing Muse Wonder and Muse Wonder Full, described as its most capable models to date for personal and professional agentic tasks. The models are built on the groundwork laid with Muse Spark. Fourteen months ago, Meta began an overhaul of its AI efforts, according to the announcement.

The new models are aimed at expert domains including law, finance, and health, and at frontier coding performance for people building everything from a game to an app for a new business. They include Active Memory, which carries what matters to you across conversations and brings back relevant details, and Active Focus, which keeps working toward your goal when a task gets hard, tries other approaches, recovers from errors, and follows through, asking for clarification when needed and checking in before following through on anything important.

Meta says the Muse Wonder family was trained to avoid sharing private information without permission, turn down requests to assist in performing actually harmful acts, and work to improve security and strengthen defenses against cyber attacks. Developers and security teams can still use models like Muse Wonder to find flaws in their own code and strengthen their systems. Muse Wonder is available today through the Meta Model API and Muse Code, and will soon power more of Muse, rolling out first to subscribers. A full safety report will be available soon.

Key facts

• Muse Wonder and Muse Wonder Full are Meta’s most capable models to date, built on the groundwork laid with Muse Spark.
• Active Memory carries relevant details across conversations; Active Focus keeps working toward a goal, recovers from errors, asks for clarification, and checks in before important actions.
• The models target expert domains including law, finance, and health, plus frontier coding performance for people building from a game to an app for a new business.
• Muse Wonder is available today through the Meta Model API and Muse Code, and will soon power more of Muse, rolling out first to subscribers.
• Safety training includes avoiding sharing private information without permission, refusing actually harmful acts, and strengthening defenses against cyber attacks.

Our analysis

The launch signals a shift from one-off AI responses to persistent agents that can work over time and across messages, calendars, and documents. For social media managers and creators, that likely reduces the manual work of re-establishing context for recurring campaigns, support threads, or long-term content projects. The emphasis on Active Focus also suggests Meta is targeting tasks where earlier models stalled: multi-step problems, error recovery, and confirmation before high-stakes actions.

What is different from before is the combination of expert-domain capability with memory and focus features. Rather than only generating text or code, Muse Wonder is positioned to interpret information and work through a problem, which matters for operators who manage finance, legal, or health-adjacent content and need more reliable outputs. The privacy guardrails suggest Meta expects these agents to operate close to sensitive personal data, so permission boundaries will likely determine whether brand accounts and community teams can safely delegate access.

The fact that developers can still use the models to find flaws in their own code is a meaningful detail: it suggests safety restrictions are designed to block harmful acts while preserving legitimate defensive and coding work. Because availability starts through the API and Muse Code before a subscriber rollout, operators can evaluate capabilities early rather than waiting for the consumer version.

What it means for operators

• Test the Meta Model API and Muse Code now for multi-step automation, coding help, and content workflows that require memory across sessions.
• Before connecting calendars, messages, or documents, set explicit permission rules and review checkpoints for anything the agent proposes to do on your behalf.
• Prepare for the subscriber rollout by listing long-running tasks that need Active Focus, such as campaign troubleshooting, bot debugging, or managing frequent creator negotiations.
• Track the upcoming full safety report and the research blog before updating brand AI policies or building customer-facing community bots.