The research report released by CITIC Securities highlights that the domain of embodied AI large models is currently undergoing a wave of frequent updates. The adoption of UMI data collection has become widespread, primarily owing to its distinctive features of "low cost and high efficiency." Moreover, the large-scale utilization of first-person video data has garnered provisional endorsement from leading industry players. It is anticipated that the abundant and varied video learning approaches will serve as a catalyst for the advancement of world models, leveraging the foundation laid by video generation models. The year 2026 is poised to be a landmark year, marking the commencement of mass production within the embodied intelligent robot sector. Presently, the sophistication of embodied models trails behind that of their hardware counterparts, underscoring the paramount importance of their iterative evolution. In parallel, as firms like Skild AI, Physical Intelligence, and Galbot progress through their capitalization endeavors, it is expected that the pricing rationale for companies specializing in embodied models will crystallize into a clearer and more standardized framework.