Odyssey AI released Odyssey-3, a generative world model, on October 11, 2026, saying it generates interactive environments from text prompts in real time. It is the company's first major update since its public research preview on September 15, 2026.
Odyssey AI reported 66.1 points on the video-to-video benchmark from Physics-IQ Verified, measured on a test evaluating physical behavior across areas like fluid mechanics, optics, solid mechanics, magnetism, and thermodynamics. That compares with 63.37 points on the same benchmark without a selection method.
Odyssey-3 is built on an autoregressive diffusion transformer and targets robotics, autonomous driving, and video games. Availability begins with a public research preview, initially for developers applying for API access.
"The model learns physical relationships and cause-and-effect from visual observations during training," said Oliver Cameron, co-founder of Odyssey AI. The model uses training data including internet videos with event descriptions, video game footage paired with the corresponding keyboard and mouse inputs, and simulated physical interactions.
The announcement follows the public release of the model's technical details and benchmark results. Odyssey AI said the model is designed to serve as a foundation for controlling different systems, from robotic arms and drones in simulated environments to characters in video games.
Odyssey AI did not say how the model handles edge cases, and it raised the open question of whether the top score meets benchmark rules requiring four test runs with standard deviation reported. The public research preview offers interactive environment generation for now, robotics and autonomous systems will need further work.
Source: thedecoder
