MENLO PARK, CALIFORNIA / RankWire.AI / – Meta announced the release of Muse Wonder and Muse Wonder Full, a new series of sophisticated artificial intelligence models designed for software development, AI agents, and complex everyday tasks. These systems are capable of assisting with areas such as leases, insurance claims, financial information, and application development. On October 12, Meta launched Muse Wonder via its Model API and Muse Code, making Muse Wonder Full accessible to users with eligible subscriptions.
Focusing on multi-step tasks rather than single responses, the models enable planning, digital tool integration, and task division among specialized sub-agents. They are capable of functioning across browsers, desktop applications, and mobile platforms. Designed to recover from failed approaches, the system supports context windows of approximately one million tokens, allowing it to process large documents, extensive codebases, and other sizable materials within a single session.
Meta also introduced Active Memory and Active Focus features alongside the Wonder models. Active Memory retains useful information from prior conversations and retrieves it as needed. Active Focus assists AI agents in navigating difficult problems and adjusting their strategies after errors. The models can ask for clarification when instructions lack critical details and seek confirmation before executing sensitive or significant actions. These tools aim to facilitate longer work sessions without requiring users to restart tasks repeatedly.
Meta’s new models expand agent and coding functionalities
Pricing for Muse Wonder is set at $4 per million input tokens via the Meta Model API. Output tokens cost $20 per million, while cached input tokens are priced at $0.20 per million. Instead of separate API charges, Meta offers Muse Wonder Full through paid Muse Code plans, with a Power Usage plan costing $50 monthly. For heavier coding needs, the Maximum Usage plan is priced at $200 per month, targeting developers requiring greater access.
Meta shared benchmark results assessing software engineering, coding agents, and lengthy technical tasks. The tests included DeepSWE and SWE-Marathon, with evaluations by Artificial Analysis at maximum reasoning effort. Muse Wonder received an Intelligence Index score of 52, while Muse Wonder Full scored 56. These scores surpass those of Meta’s earlier Muse Spark models, indicating a clear improvement in reasoning capabilities on the evaluator’s scale.
Safety and privacy measures accompany the rollout
Meta states that the Wonder models are equipped with safeguards for privacy, cybersecurity, and harmful request prevention. They are trained to avoid revealing private information without user consent and to reject 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 plans to share further details about safety testing for these new models.
This launch follows a significant reorganization of Meta’s artificial intelligence initiatives that started roughly 14 months earlier. The company states that Muse Wonder builds upon the technical foundations established by Muse Spark. Development of the new models utilized computing resources at Meta’s Prometheus data center. Additionally, the company intends to integrate Muse Wonder technology into more features within the Muse platform, initially targeting subscribers. Developers can already access Muse Wonder, while Muse Wonder Full remains available through qualifying Muse Code subscription plans.