Meta's new Muse Wonder and Muse Wonder Full models achieve frontier performance, but only the weaker one is available via API
Meta is back in the AI game with two new proprietary models, Muse Wonder and Muse Wonder Full, which the company claims can outperform leading offerings from OpenAI and Anthropic on complex tasks such as software engineering, legal work, and business automation.
The models are designed for long-running agent tasks, such as software engineering, legal work, and business automation, and feature active memory and active focus capabilities, allowing agents to navigate multiple applications, retain information, and recover from problems without abandoning the user's objective.
Meta is making the standard Wonder model available through a commercial API, but Wonder Full is reserved for paid subscribers to its Muse Code programming assistant, which means enterprises will need to weigh the benefits of access to this powerful model against the costs and limitations of Meta's commercial terms.
Story 1 of 10