Meta launches Muse Wonder and Wonder Full, new AI models for personal and professional tasks

On Monday, 12th, Meta announced two new artificial intelligence models: Muse Wonder and Muse Wonder Full. According to the company, these models are designed to perform a wide range of tasks, from document processing to application development, catering to both personal and professional needs.

Availability and integration

Muse Wonder can be accessed via the Meta Model API, which enables integration with other applications, as well as through Muse Code, the company’s programming tool. Muse Wonder Full, however, is exclusively available to Muse Code subscribers. Meta plans to integrate Wonder into the Muse personal assistant in a future update, starting with subscribers of the service.

Difference between Muse and Muse Wonder

Muse is a personal assistant product, while Muse Wonder is an AI model capable of performing some of the tasks handled by this assistant. Wonder Full offers additional exclusive features for Muse Code subscribers.

Features and applications

Meta highlights two key features: Active Memory, which retains relevant information between conversations and retrieves it when needed, and Active Focus, which enables the system to persist in a task when encountering difficulties, trying alternative approaches, correcting errors, and seeking clarification.

Use cases include interpreting rental contracts, processing insurance claims, and organizing financial information. These examples are illustrative and do not represent an independent assessment of accuracy.

Programming and persistence

In programming, the models have been trained to handle long-term projects that require tracking requirements, modifying files, and repeating tests. The goal is to support software development, not just complete isolated code snippets.

Competitive landscape

This launch marks another round in Meta’s competition with OpenAI and Anthropic. The company claims that the Muse family is the most capable it has ever developed, with a focus on programming, memory, and tasks requiring persistence across multiple stages. Meta publishes performance comparisons and claims top results, although its own tests do not prove superiority in any specific task.

The company says a full report with additional security assessments will be released later. Integration with Muse will serve as another test of personal assistants’ ability to maintain context and continue work without losing sight of the original request.