MENLO PARK, CALIFORNIA / RankWire.AI / – Meta has unveiled Muse Wonder and Muse Wonder Full, a new set of sophisticated artificial intelligence models. These systems are designed for software development, AI agents, and tackling complex everyday tasks. They are capable of assisting with matters such as leases, insurance claims, financial data, and application building. Meta launched Muse Wonder via its Model API and Muse Code on October 12. The more comprehensive Muse Wonder Full became accessible to users with eligible Muse Code subscriptions.
The focus of these models is on multi-step tasks rather than single-answer queries. Muse Wonder can strategize workflows, utilize digital tools, and distribute larger jobs among specialized sub-agents. Compatibility across browsers, desktop software, and mobile apps is also a feature. Designed for resilience, the system can recover from failures of a chosen approach. Supporting context windows of approximately one million tokens, the models can process large documents, extensive codebases, and other substantial materials in a single session.
Meta has also introduced Active Memory and Active Focus alongside the Wonder models. Active Memory retains valuable information from previous conversations and retrieves it when necessary. Active Focus enables an AI agent to persist through difficult problems and adapt its strategy after mistakes. The models can ask for clarification if instructions lack critical details and seek confirmation before executing sensitive or significant actions. These features aim to facilitate longer work sessions without requiring users to restart tasks repeatedly.
Expanded agent and coding features in new AI models
Pricing for Muse Wonder is set at $4 per million input tokens through the Meta Model API. Output costs $20 per million tokens, with cached input costing $0.20 per million tokens. Instead of separate API charges, Meta offers Muse Wonder Full through paid Muse Code plans. The company also provides a Power Usage plan at $50 per month, and a Maximum Usage plan costing $200 monthly, aimed at developers with heavier coding requirements.
Meta has published benchmark results that evaluate software engineering, coding agents, and extended technical tasks. Key tests include DeepSWE and SWE-Marathon. The evaluation also involved maximum reasoning effort tests, with Muse Wonder earning an Intelligence Index score of 52 and Muse Wonder Full scoring 56. These results surpass Meta’s earlier Muse Spark models on the evaluator’s scale, indicating a clear improvement in reasoning performance.
Safety protocols and privacy safeguards included in the launch
Meta states that the Wonder models incorporate protections against privacy breaches, harmful requests, and cybersecurity threats. They are trained to refrain from revealing private information without user consent and to decline harmful activity requests, while still supporting legitimate security functions. Developers can utilize the models to identify vulnerabilities in their software and enhance defenses. 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 release follows a restructuring of Meta’s AI efforts that started roughly 14 months earlier. The company explains that Muse Wonder builds upon the technical foundation laid by Muse Spark. Development utilized computing resources at Meta’s Prometheus data center. Plans are underway to integrate Muse Wonder technology into additional features within the Muse ecosystem, initially targeting subscribers. While developers can already access Muse Wonder, Muse Wonder Full remains available through qualifying Muse Code subscription plans.