Meta launches Muse Wonder, its most capable AI models for agents and coding

Meta has introduced the Muse Wonder series — Muse Wonder and Muse Wonder Full — which the company describes as its most capable models yet, aimed at personal and professional "agentic" tasks: AI that plans and carries out multi-step work rather than just answering questions.

Two models, two ways to access them

• Muse Wonder is available now through the Meta Model API and Muse Code, Meta's coding tool. API pricing is $4 per million input tokens and $20 per million output tokens, with cached input at $0.20 per million. Meta says it offers performance comparable to the Full model.
• Muse Wonder Full is available only by subscription in Muse Code, on the Power Usage plan and a new Maximum Usage plan. Meta calls it its most advanced model, with a focus on coding, long-running agents and computer use.

The new Maximum Usage plan includes 80 times the limits of the Everyday Usage plan, with the same five-hour and weekly resets.

What Meta says the models can do

According to Meta, Muse Wonder can plan long tasks, delegate work to sub-agents and recover from errors. The company highlights two features it calls Active Memory and Active Focus, intended to keep track of what matters to the user and stay on task.

Meta says the models were trained at its gigawatt-scale Prometheus data center and build on the scaling approach it described with the earlier Muse Spark model.

What's next

Meta plans to use Muse Wonder to power experiences in its Muse assistant, rolling out first to subscribers, and to develop specialized models for wider use later.

Keep in mind

Performance claims in launch announcements come from the company itself. Independent evaluations typically follow in the days and weeks after a release; if you're choosing a model for real work, test it on your own tasks and compare pricing against alternatives.