The AI race is shifting toward “world models,” a fast-emerging technology designed to understand and simulate physical reality. Supporters say these systems could power breakthroughs in robotics, gaming, manufacturing and autonomous tools that require real-world reasoning, something large language models still struggle to deliver.
Leading researchers are moving aggressively. Fei-Fei Li’s World Labs has released its first commercial model, Marble. Yann LeCun is preparing to launch a world-model startup after leaving Meta, and both Google and Meta are developing their own systems.
AI's next big leap is models that understand the world. https://t.co/ImMCfruhco
— Axios (@axios) November 17, 2025
OpenAI argues that improved video models may be a pathway toward full world modeling. Globally, China’s Tencent and the UAE’s Mohamed bin Zayed University are investing heavily.
World models learn from video, 3D data and simulations, enabling them to predict physical behavior instead of the next word.
The biggest challenge is data scarcity. Developers say production-grade systems will require far more high-quality multimodal data than the internet currently provides.
Also read:
TIPP Staff
