Odyssey said on October 8 that it launched Odyssey-3, a world model for generating navigable simulated environments, alongside a browser-based research preview.
According to the Odyssey announcement, the system is built to keep a scene updating as a person or agent moves, acts, or introduces new events. The preview supports first-person and third-person navigation plus independent camera movement.
Odyssey lists two versions: Odyssey-3 at 832×480 and Odyssey-3 Pro at 1280×720. The company also says the Pro model scored 66.1 on Physics-IQ Verified video-to-video, a vendor-reported benchmark claim that still lacks broader third-party validation.
For teams working on embodied agents, the near-term point is evaluation rather than broad deployment. The browser preview offers a way to test Odyssey’s simulation approach for agent training or environment-centric workflows, while building access for physical AI projects remains request-based and Odyssey has not posted public pricing or open weights.