California AI startup Odyssey has opened a public research preview of its world model, Odyssey‑3. Anyone can try the free online demo, which creates interactive environments from a text description and lets users move in first‑person or third‑person view. Developers can also apply for API access to integrate the model into their own projects.
What the model can do
Odyssey‑3 builds video frames in real time using an autoregressive diffusion transformer, a type of AI that predicts each frame from the previous one and from user actions. The base model has 14 billion parameters and outputs video at 832 × 480 pixels; a higher‑tier version, Odyssey‑3 Pro, runs at 1280 × 720 pixels. The company says the model learns cause‑and‑effect from internet videos, game footage with input logs, and simulated physics scenes.
Benchmarks and early tests
Odyssey reports a top score of 66.1 points on the Physics‑IQ Verified video‑to‑video benchmark, though the record comes from a single run with a selection step that does not meet the benchmark’s standard rules. Without that step, the model averaged 63.37 points over four runs. On the WorldMark benchmark, Odyssey‑3 placed first in three of four categories, scoring between 76.3 and 79.0 points. The company also demoed the model controlling robotic arms, drones, and game characters, claiming better reliability than competing systems.
Why it matters
For hobbyists and developers, the free demo offers a hands‑on way to explore AI‑generated 3‑D worlds without needing a high‑end GPU. Companies can apply for API access to prototype robotics or game‑AI concepts, though the model’s benchmark claims still need independent verification.
