Meta has launched V-JEPA 2, an advanced AI world model designed to enhance robots' and autonomous vehicles' ability to understand and interact with the physical environment. This 1.2 billion parameter model learns from raw video data without requiring labels, enabling it to predict actions and consequences, perform zero-shot generalization to new tasks, and plan and execute operations in unfamiliar settings. V-JEPA 2 is trained on over one million hours of internet video combined with robot interaction data, allowing it to perceive, understand, and respond to real-world scenarios with human-like reasoning. Meta claims that V-JEPA 2 predicts real-world interactions and actions up to 30 times faster than Nvidia's Cosmos model, marking a notable advancement in robotics and autonomous systems. The model is open-source and aims to accelerate physical AI development, particularly benefiting delivery robots, self-driving cars, and other intelligent agents that require real-time navigation and decision-making capabilities.