Meta, the parent company of Facebook and Instagram announced on Wednesday that it is rolling out a new AI “world model” capable of understanding 3D environments and the movement of physical objects more effectively.

Meta said it launched V-JEPA 2, an open-source AI model designed to understand, predict, and plan in the physical world.

As a “world model,” V-JEPA 2 draws on principles of real-world logic to create internal simulations, allowing AI to learn, reason, and make decisions more like humans.

AI has emerged as a top priority for Meta CEO Mark Zuckerberg, as the company competes with major tech firms such as OpenAI, Microsoft, and Google.

Meta showcased the capabilities of its V-JEPA 2 model in applications such as delivery robots and self-driving cars, which need to process their environments in real time to navigate effectively.

The company explained instead of relying on vast amounts of labeled data or raw video, V-JEPA 2 uses a simplified “latent” space to reason about object movement, interaction, and response,

World models have been gaining significant attention in the AI community as researchers shift focus beyond large language models—the backbone of generative AI tools like OpenAI’s ChatGPT and Google’s Gemini.