What Odyssey does
Odyssey (website currently hosted on odyssey.ml and reachable at odyssey.world) is an AI lab focused on “general world models” intended to understand and simulate the world over time and under interaction—ultimately enabling long-running, interactive applications in areas like robotics, gaming, education, healthcare, and defense. The company’s core product direction is interactive video/world simulation: models that generate streams of visual (and in some cases audio) content that respond to user inputs, with an emphasis on real-time generation and longer-horizon coherence.
More
Odyssey’s flagship model line includes Odyssey-2 (and newer variants such as Odyssey-2 Pro and Odyssey-2 Max), positioned as general-purpose world models for open-ended inputs and improved physical fidelity. In parallel, Odyssey is publishing specialized research releases that expand modality and multi-agent interaction: Starchild-1 (audio-visual, real-time), Agora-1 (multi-agent shared world interaction), and PROWL-1 (reinforcement learning via adversarial exploration intended to improve world model performance). Odyssey also describes an API initiative intended to let developers integrate these interactive simulations into applications, including through what it calls “Odyssey-2 Pro” availability via API. Business model signals (limited to public sources found): Odyssey appears to be building developer access and integration paths around its models (via an API and interactive “try” demos), suggesting an enterprise/developer access model rather than a consumer marketplace. The company’s positioning emphasizes simulation quality, physical accuracy benchmarking, interactive conditioning, and scaling training/inference for real-time performance.