Meta is rolling out a series of specialized models—trained on its frontier large model, Muse Wonder—to Muse and Meta AI users for everyday conversations and complex task handling. The new generation of Muse models has improved its score on the Personal Superintelligence Index test from 38% in the previous generation to 47%, and introduces two new features: proactive memory and proactive focus. Proactive memory retains key details across months of conversations, achieving a 90% accuracy rate in recalling key context and greatly reducing the error rate of forgetting critical information. Proactive focus enables the model to attempt recovery or offer alternatives when a task encounters unexpected situations, boosting performance on related tasks by 67%. In addition, the models have been finely tuned to the distinct scenarios of the two products: on Muse, the model can independently advance tasks or collaborate with users, while on Meta AI, it can break down complex problems, ask precise follow-up questions, and smoothly adapt to the direction of the user's conversation—allowing ordinary users to experience superintelligence with long-lasting memory and autonomous task-focus capabilities.