MENLO PARK, CALIFORNIA / RankWire.AI / – Meta has introduced Muse Wonder and Muse Wonder Full, a new lineup of sophisticated artificial intelligence models. These systems are crafted for software development, AI agents, and handling complex daily tasks. They assist with activities such as leases, insurance claims, financial data, and app creation. Meta launched Muse Wonder via its Model API and Muse Code on October 12. Muse Wonder Full is now accessible to users with eligible Muse Code subscriptions.
Designed for multi-step tasks rather than simple responses, the models excel at planning, leveraging digital tools, and splitting large projects among specialized sub-agents. They can operate across browsers, desktop software, and mobile apps. Meta built the system to recover when an approach fails. These models support context windows of approximately one million tokens, enabling them to process extensive documents, codebases, and large materials in a single session.
Meta also launched Active Memory and Active Focus alongside the Wonder models. Active Memory retains useful information from previous conversations and retrieves it when necessary. Active Focus helps AI agents persist through difficult problems and adapt their strategies after errors. The models can ask for clarification if instructions lack key details. They can also confirm before executing sensitive or critical actions. These features aim to support longer work periods without users needing to restart tasks repeatedly.
Enhanced capabilities for coding and agent functions
Muse Wonder is priced at $4 per million input tokens via the Meta Model API. Output costs $20 per million tokens, while cached input costs $0.20 per million tokens. Instead of separate API pricing, Meta offers Muse Wonder Full through paid Muse Code plans. The company provides a Power Usage plan at $50 per month, and its Maximum Usage plan costs $200 monthly. The latter targets developers requiring more extensive access for intensive coding tasks.
Meta shared benchmark results for software engineering, coding agents, and long-term technical tasks. They highlighted tests like DeepSWE and SWE-Marathon. Additionally, Artificial Analysis assessed both models 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 and indicate improved reasoning performance based on evaluator metrics.
Safety measures and privacy protections included in launch
Meta states that the Wonder models incorporate safeguards for privacy, harmful requests, and cybersecurity. They are trained to avoid revealing private information without user approval. The models also reject requests involving harmful activities while maintaining legitimate security functions. Developers can still use these models to identify vulnerabilities in their software and enhance security measures. Meta is assessing its advanced AI systems under its Superintelligence Scaling Framework. The company plans to publish further details about safety testing for the new models.
This launch follows a significant restructuring of Meta’s AI efforts that began roughly 14 months ago. The company states Muse Wonder builds upon foundational work developed through Muse Spark. The new models were trained using computing resources at Meta’s Prometheus data center. Additionally, Meta intends to integrate Muse Wonder into more features within the Muse platform, starting with subscribers. Developers can already access Muse Wonder, while Muse Wonder Full remains available through qualifying Muse Code subscriptions.