Odyssey-3 is a world model you can act inside

Watch

Preview

Odyssey released a world model you can move inside, with a public research preview.

How to read this claim

Measurement — Physics-IQ and the robot results are Odyssey's. They have not been independently rerun.
Bill — List input and output per million tokens on the vendor page — a loop pays both again.
Door — Physics-IQ and the robot results are Odyssey's. They have not been independently rerun.

Keep the habit · field card with prices and worked checks

What changed

A video model shows a scene. Odyssey-3 claims a scene you can act in, plus a small policy on a frozen world model. That second claim is the lab's, and there is no robot to deploy from the post.

What it introduces

A public preview of a scene you can act inside. It is not a robot or a driving system you can deploy from the post.

Who benefits Robotics and autonomy teams, simulation groups, and anyone comparing world models as a place to train agents. API-style access for physical-AI developers is a contact path, not a general launch.

Try this

Watch this. Do not plan a robot or vehicle deployment from the preview.

What we do not know

Physics-IQ and the robot results are Odyssey's. They have not been independently rerun.

Evidence

Primary Odyssey. Scores and robot results reported by Odyssey; not independently verified.

Independent Physics-IQ and the robot results are Odyssey's. They have not been independently rerun.

Sources

Odyssey Primary · October 8, 2026

Still unclear? · Start here · ranks, prices, unknown Physics-IQ and the robot results are Odyssey's. They have not been independently rerun.

The record

What happened

On October 8, 2026, Oliver Cameron and Jeff Hawke published Odyssey-3. They describe an autoregressive diffusion transformer trained to predict motion, contact, and what happens next. The research preview supports first-person and third-person movement and an independent camera. Odyssey-3 Pro scores 66.1 on Physics-IQ Verified’s video-to-video benchmark, which Odyssey calls the highest reported score as of October 7, and 54.7 on image-to-video.

Why it matters

A world model is a learned guess about how a scene evolves. Odyssey’s claim is that the same backbone can take a thin action decoder and drive different bodies: arms, humanoids, a car. A score for predicting a video of a physics experiment is not a policy you can ship. Odyssey says a frozen model plus a small policy still behaved when the scene changed. That result needs outside replication. Until then this is a serious preview, not a robot.

Odyssey reports robot-arm tasks from tens of hours of demonstrations, including recoveries that were not in those demonstrations, such as reorienting a gripper after a miss. Flexion, they say, trained humanoid policies on Odyssey-3 that beat the VLA baselines they tested when the environment and the lighting changed. A driving policy trained on 20 hours of road data in India, with the backbone frozen, predicts waypoints and drives in closed loop. An early checkpoint also generated three camera views together after 100 training steps.

The model is trained as a multi-step video diffusion transformer, extended autoregressively with teacher forcing and causal masking so it continues from what it has already generated, conditioned on actions. A distillation pass produces a few-step variant for real-time interaction. The preview lets a person prompt a world, move, and introduce an event while it is generating. Training data, as described, mixes internet video with event labels, gameplay with keyboard and mouse, and simulated rigid-body interactions.

WorldMark, in Odyssey’s evaluation, averages 13 metrics of control-following, visual quality, and memory. The lab says the model ranks first in three of four categories in its own run: first-person stylized, third-person real (79.0), and third-person stylized (76.3). Physics-IQ continues videos of real experiments in fluids, optics, solid mechanics, magnetism, and thermodynamics. Both leaderboards are Odyssey’s report.