88Z AI, 2 September — World Labs, founded by Li Fei-Fei, has unveiled Atlas, the world's first multimodal world model. Built on a multimodal autoregressive diffusion Transformer architecture, it features pixel-level camera control, 3D reconstruction, and spatiotemporal simulation. Outperforming current state-of-the-art models, Atlas is now available for early access to select partners.
At a glance
- Atlas is the world's first multimodal world model, surpassing current SOTA performance
- Supports pixel-level camera control, 3D reconstruction, and spatiotemporal simulation with early access available
The update
World Labs, led by Li Fei-Fei, has officially launched Atlas, the world's first multimodal world model. Utilizing an innovative multimodal autoregressive diffusion Transformer architecture, its core capabilities include image and video generation with pixel-level camera control, high-precision 3D scene reconstruction, and complex spatiotemporal simulation. Atlas outperforms existing state-of-the-art models in key tasks such as camera-controlled generation and 3D reconstruction. It is regarded as a critical step forward for Real-to-Sim technology in robotics. Currently, World Labs is offering early access to Atlas for select partners to accelerate application development and exploration.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 量子位Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.