MENLO PARK, CALIFORNIA / RankWire.AI / – Meta has introduced the Muse Wonder and Muse Wonder Full models, a new lineup of sophisticated artificial intelligence systems. These models are designed to support software development, AI agents, and complex daily tasks. They assist with activities such as managing leases, processing insurance claims, handling financial data, and developing applications. Meta launched Muse Wonder via its Model API and Muse Code on October 12, with Muse Wonder Full becoming accessible to users with eligible Muse Code subscriptions.
The focus of these models is on multi-step tasks rather than single responses. Muse Wonder can strategize, utilize digital tools, and break down larger projects into smaller, specialized sub-agents. It is capable of functioning across web browsers, desktop applications, and mobile platforms. Meta designed the system to recover gracefully when an initial approach fails. The models support context windows of approximately one million tokens, enabling them to process large documents, extensive codebases, and other sizable data within a single session.
Meta also rolled out Active Memory and Active Focus features alongside the Wonder models. Active Memory retains valuable details from previous conversations and retrieves them as needed. Active Focus helps an AI agent persist through challenging problems and adapt its approach following errors. The models can ask for clarification if instructions lack critical information and seek approval before executing sensitive or significant actions. These enhancements aim to facilitate longer work sessions without requiring users to restart their tasks repeatedly.
New models boost Meta’s agent and programming capabilities
The cost of Muse Wonder is $4 per million input tokens via the Meta Model API. Output tokens are priced at $20 per million, while cached input tokens cost $0.20 per million. Instead of separate API charges, Meta offers Muse Wonder Full through paid Muse Code plans. The company provides a Power Usage plan at $50 monthly and a Maximum Usage plan at $200 per month, catering to developers with higher demands for intensive coding tasks.
Meta shared benchmark results related to software engineering, coding agents, and long-duration technical tasks. Tests such as DeepSWE and SWE-Marathon contributed to these results. Additionally, Artificial Analysis assessed both models at maximum reasoning effort, awarding Muse Wonder an Intelligence Index of 52 and Muse Wonder Full a score of 56. These scores surpass Meta’s previous Muse Spark models, indicating a clear improvement in reasoning performance based on the evaluator’s scale.
Safety and privacy measures accompany the rollout
Meta emphasizes that the Wonder models incorporate safeguards for privacy, harmful requests, and cybersecurity. They are trained to avoid revealing private information without user consent and to decline requests involving malicious activities while maintaining essential security functions. Developers can utilize these models to identify vulnerabilities in their own software and strengthen defenses. Meta assesses its most advanced AI systems under its Superintelligence Scaling Framework and plans to release additional information about safety testing for the new models.
This release follows a comprehensive restructuring of Meta’s AI initiatives that began approximately 14 months ago. The company states that Muse Wonder builds upon the technical foundation laid by Muse Spark. Development utilized computing resources at its Prometheus data center. Meta also intends to integrate Muse Wonder technology into more features within the Muse ecosystem, starting with subscribers. While developers can already access Muse Wonder, Muse Wonder Full remains available through qualifying Muse Code subscription plans.