Back

 Industry News Details

 
Meta's V-JEPA 2 Empowers AI to Perceive and Navigate the Physical World Posted on : Jun 12 - 2025
Meta has introduced its latest AI model, V-JEPA 2, a next-generation “world model” designed to help AI agents better understand and navigate the physical world around them.
 
V-JEPA 2 builds on last year’s V-JEPA, which was trained on more than 1 million hours of video to teach AI systems how real-world dynamics—like gravity or motion—unfold over time. This kind of training enables AI agents, such as robots, to make predictions and decisions with a level of common-sense reasoning similar to that of young children or animals. For instance, a dog chasing a ball instinctively predicts its trajectory rather than simply following where it is at the moment—a type of intuitive learning V-JEPA 2 aims to replicate in machines.
 
Meta showcased examples like a robot holding a spatula and plate while approaching a stove. V-JEPA 2 can infer that the most probable next step is to transfer the cooked eggs to the plate—demonstrating its grasp of goal-oriented actions.
 
According to Meta, V-JEPA 2 outperforms similar systems, claiming to be 30 times faster than Nvidia’s Cosmos model, which also focuses on physical-world intelligence. However, these performance comparisons may rely on differing benchmarks.
 
Meta’s Chief AI Scientist, Yann LeCun, emphasized the transformative potential of world models:
 
“We believe world models will usher in a new era for robotics, enabling real-world AI agents to help with chores and physical tasks without needing astronomical amounts of robotic training data.”