World models are real research. Ask vendors to show closed-loop control.
Meta's V-JEPA 2 and Google DeepMind's Genie 3 show genuine progress. Neither is a turnkey product for vehicles or yards.
Meta released V-JEPA 2 in June 2025, a video world model pretrained on more than a million hours of video and then fine-tuned on 62 hours of robot data. Meta reported 65 to 80 percent success on pick-and-place tasks with new objects and settings.
Google DeepMind announced Genie 3 in August 2025. It generates interactive environments in real time at 720p and 24 frames per second that stay consistent for a few minutes, and was released as a limited research preview.
ESS view: these are research milestones. Agencies evaluating yard automation or shuttles marketed as using world models should ask to see closed-loop control in their own environment, not generated video.