MENLO PARK, CALIFORNIA / RankWire.AI / – Meta has unveiled Muse Wonder and Muse Wonder Full, a new lineup of sophisticated artificial intelligence models. These systems are designed for software development, AI agents, and handling complex everyday tasks. They can assist with activities such as managing leases, processing insurance claims, accessing financial data, and developing applications. Meta launched Muse Wonder through its Model API and Muse Code on October 12, with Muse Wonder Full becoming accessible to users with eligible Muse Code subscriptions.
Designed for multi-step tasks rather than single-answer queries, the models excel at planning work, utilizing digital tools, and breaking down large projects into specialized sub-agents. They operate seamlessly across browsers, desktop applications, and mobile devices. Meta engineered the system to recover effectively if a chosen approach fails. Supporting context windows of approximately one million tokens, the models can analyze large documents, extensive codebases, and other substantial materials within a single session.
Meta also introduced features called Active Memory and Active Focus with the Wonder models. Active Memory retains useful information from previous conversations and retrieves it as needed. Active Focus allows an AI agent to persist through difficult problems and adapt its approach after encountering errors. The models can ask for clarification when instructions lack vital details and seek confirmation before executing sensitive or critical actions. These enhancements aim to enable longer work sessions without requiring users to restart tasks repeatedly.
Enhanced agent and coding features introduced with new models
The cost of Muse Wonder is $4 per million input tokens via the Meta Model API, with output priced at $20 per million tokens. Cached input costs $0.20 per million tokens. Instead of separate API pricing, Meta offers Muse Wonder Full through paid Muse Code plans, including a Power Usage plan at $50 per month and a Maximum Usage plan at $200 per month, aimed at developers requiring more extensive access for intensive coding tasks.
Meta shared benchmark results covering areas like software engineering, coding agents, and long-duration technical operations. The tests included DeepSWE and SWE-Marathon, with the models also evaluated at maximum reasoning effort. Muse Wonder received an Intelligence Index score of 52, while Muse Wonder Full scored 56. These results surpass Meta’s earlier Muse Spark models on the evaluator’s scale, indicating a notable improvement in reasoning capability.
Safety protocols and privacy safeguards added with the launch
Meta states that the Wonder models incorporate protections to address privacy, harmful requests, and cybersecurity concerns. They are trained to prevent disclosing private information without user consent and to decline requests involving harmful activities, while still supporting legitimate security functions. Developers can utilize the models to identify vulnerabilities in their own software and enhance security measures. Meta is assessing its most advanced AI systems under its Superintelligence Scaling Framework and intends to publish further details about safety testing for these new models.
The rollout follows a restructuring of Meta’s AI initiatives that began approximately 14 months ago. The company notes that Muse Wonder builds upon the technical foundation laid by Muse Spark, with training conducted at its Prometheus data center. Plans are underway to incorporate Muse Wonder technology into additional features within the Muse platform, starting with subscribers. Developers can access Muse Wonder now, while Muse Wonder Full remains available through qualifying Muse Code subscription plans.