Daxiao Robotics and NTU S-Lab Open-Source Unified World Model Puffin-World

On September 10, 2026, Daxiao Robotics, together with NTU's S-Lab and other institutions, released and open-sourced Puffin-World, a unified multimodal world model framework.

The framework treats physics, geometry, and appearance as three native 3D world states, enabling camera pose estimation, free-viewpoint rendering, 3D reconstruction, and closed-loop robot exploration. The team also open-sourced the code, models, and the Puffin-16M dataset, which contains 15 million vision-language-camera data pairs, 1 million rotation trajectories, and absolute camera pose annotations for approximately 44.5 million images across 28 public datasets.