Odyssey
About
About Odyssey
Odyssey builds general world models that learn physical dynamics, causality and interaction from multimodal data. Odyssey-2 generates long-running interactive simulations in real time; Starchild-1 incorporates multimodal interaction; Agora-1 supports multiple agents in a shared simulated environment; and PROWL-1 uses reinforcement learning to explore and improve world-model performance.
The company positions these models as simulation infrastructure for robots and other embodied agents, including policy evaluation and controllable edge-case generation before actions are taken in the physical world. Odyssey was founded in 2023 by former autonomous-driving executives Oliver Cameron and Jeff Hawke and operates from Palo Alto, London and Zurich.
Articles
Funding
- 2026-06-17 - Series B - $310M - Odyssey raises $310M Series B at stated $1.45B valuation
Odyssey completed a private financing to accelerate its world-model research and deployment. - source: odyssey.ml
Investors
- Amazon - Participant - Corporate / CVC - 1 funding event - strategic/corporate - first seen 2026-06-17
- AMD Ventures - Participant - VC - 1 funding event - first seen 2026-06-17
- EQT - Participant - Unknown - 1 funding event - first seen 2026-06-17
- GV - Participant - Unknown - 1 funding event - first seen 2026-06-17
- IQT - Participant - Unknown - 1 funding event - first seen 2026-06-17
- Natural Capital - Lead - VC - 1 funding event - first seen 2026-06-17
Team
- Oliver Cameron - Co-founder and Chief Executive Officer - Leads Odyssey and its general world-model programme.
- Jeff Hawke - Co-founder and Chief Technology Officer - Sets technical direction for Odyssey model research and systems engineering.
Research
- Odyssey releases Odyssey-3 foundation world model research preview
2026-10-08 - Research Publication - Odyssey reports Pro scores of 66.1 on Physics-IQ Verified video-to-video and 54.7 on image-to-video using prompt enhancement and best-of-eight selection. The company reports first place in three of four WorldMark categories in its own evaluations. Robot demonstrations include recovery from missed grasps and dropped objects; physical-system adaptation uses learned action decoders or policies.
Source - Odyssey releases Odyssey-3 multi-embodiment world model
2026-09-15 - Research Publication - Odyssey reports physical robot-arm skills learned from tens of hours of demonstrations and Flexion humanoid policies trained from tens of hours of teleoperation data. A driving policy trained on 20 hours of simulated data ran closed-loop on public roads in India and traveled about 77% as far between safety-driver interventions as policies trained on real footage. The drone result remains simulation-only: tens of hours of simulated flight data produced stable indoor navigation and obstacle avoidance in simulation.
Source - Odyssey releases CaliBench physical-calibration benchmark
2026-08-10 - Research Publication - The benchmark evaluates physical calibration rather than only visual plausibility. It targets model use in robotics, science, healthcare, education and gaming. The release is a research benchmark and does not establish real-world robot deployment.
Source
Timeline
Events
- 2026-10-08 - Research Publication - Odyssey releases Odyssey-3 foundation world model research preview
Odyssey reports Pro scores of 66.1 on Physics-IQ Verified video-to-video and 54.7 on image-to-video using prompt enhancement and best-of-eight selection. The company reports first place in three of four WorldMark categories in its own evaluations. Robot demonstrations include recovery from missed grasps and dropped objects; physical-system adaptation uses learned action decoders or policies.
Source - 2026-09-15 - Research Publication - Odyssey releases Odyssey-3 multi-embodiment world model
Odyssey reports physical robot-arm skills learned from tens of hours of demonstrations and Flexion humanoid policies trained from tens of hours of teleoperation data. A driving policy trained on 20 hours of simulated data ran closed-loop on public roads in India and traveled about 77% as far between safety-driver interventions as policies trained on real footage. The drone result remains simulation-only: tens of hours of simulated flight data produced stable indoor navigation and obstacle avoidance in simulation.
Source - 2026-08-10 - Research Publication - Odyssey releases CaliBench physical-calibration benchmark
The benchmark evaluates physical calibration rather than only visual plausibility. It targets model use in robotics, science, healthcare, education and gaming. The release is a research benchmark and does not establish real-world robot deployment.
Source - 2026-06-17 - Funding - Odyssey raises $310M Series B at stated $1.45B valuation
Amazon, GV, AMD Ventures, EQT and IQT participated in the financing. The funding will accelerate research and deployment of world models for robotics, autonomous systems, defense and other physical-world applications. Odyssey named AWS as its preferred cloud provider and is collaborating with Amazon Annapurna Labs on Trainium optimization, research and go-to-market work.
Source - 2026-04-21 - Product Launch - Odyssey introduces Odyssey-2 Max world model in private beta
Odyssey described three times the parameters and ten times the training compute of Odyssey-2 Pro. The model produces causal real-time rollouts that respond to actions rather than fixed-prompt video. Private-beta applications include robotics, gaming, simulation, defense and interactive systems.
Source