Odyssey-3 world model controls robots and Flexion humanoids

Odyssey-3 is the latest version of a model family Odyssey has been developing for three years, trained on a vast dataset of visual observations. Unlike models taught a specific task from scratch, Odyssey-3 provides a general understanding of physics, dynamics, cause and effect, and human behavior that can be applied across machines. The company demonstrated it across six domains, including robotics, humanoids, vehicles, AI training, drones, and games. In vehicles, it drove autonomously on roads in India using a policy trained with approximately 20 hours of simulated driving data. In humanoids, it powers control policies developed with Flexion that perform tasks in real time and generalize better than the baselines Odyssey tested. Oliver Cameron, co-founder and CEO, said world models give physical agents a foundation of knowledge they can carry from one machine, environment, or task to another. The model is designed as a foundation that spans domains rather than optimizing for a single application, which may signal a shift toward more general-purpose physical intelligence.