DINO-X World Model
A Unified Latent-Space World Model for "Generation + Understanding"
Building on DINO-X's world-leading object and action understanding capabilities, it achieves industry-leading performance in understanding physical laws and causal interactions, with the goal of bringing AI into the physical world.
Top of International Benchmarks
Visincept's world model WorldIncept topped the final WorldArena 2.0 leaderboard, ranking first worldwide in overall score on the Track 1 video generation track. It led the second-place model by 0.84 points, making it the world model with the strongest overall capabilities.
WorldArena 2.0 is one of the major international evaluation systems in the field of embodied world models. It is held in conjunction with the IROS 2026 Workshop and jointly organized by Tsinghua University, Carnegie Mellon University, Stanford University, Princeton University, and other international academic institutions. Participants across editions include leading robotics and AI teams worldwide, including Google, Nvidia, Alibaba, and Stanford University.
In addition, WorldIncept ranked first in the two core dimensions of physical consistency and 3D spatial accuracy, as well as in semantic alignment, demonstrating comprehensive leadership in understanding physical laws, spatial intelligence, and object semantics.
