World models

World models represent a growing frontier in artificial intelligence, designed to handle multiple modalities within a single architecture to advance physical AI, spatial intelligence, and robotics. Currently, the field faces challenges such as the time-consuming process of manually recording training data. To address this training data shortage, startups are exploring alternative sources, such as licensing video game controller inputs and data, though some experts warn that video game physics may not suit fine motor control tasks. In October 2026, Reka AI released a research preview of Rho-1, a 19-billion-parameter omni-model that processes text, images, video, and robot control tokens in a single shared context window. In late September 2026, General Intuition raised 220 million dollars at a 6.2 billion dollar valuation to advance its world models and hire researchers. Around the same time, AMD agreed to acquire spatial intelligence startup World Labs for 8.2 billion dollars, bringing founder Fei-Fei Li on board as chief scientist to pair AMD hardware with spatial intelligence technology. Additionally, Yann LeCun raised 1 billion dollars for his world models startup, AMI Labs, after leaving Meta earlier in the year.