World Labs, co-founded by AI pioneer Fei-Fei Li, has released Atlas, a multimodal world model capable of understanding and generating 3D spatial contexts from text, images, video, and 3D information. This model can reconstruct 3D scenes from images, generate new viewpoints, and maintain structural consistency while allowing camera movement. In parallel, Chinese AI company InSpatio has developed a similar world model, InSpatio-World, which also allows users to alter camera trajectories and explore dynamic 4D scenes, even incorporating time as an explicit dimension. InSpatio's model, InSpatio-Curious, recently topped the WorldArena 2.0 leaderboard for embodied world models, demonstrating strong performance in trajectory accuracy and physical adherence. AI
IMPACT Sets new benchmarks for spatial intelligence and 3D scene understanding, potentially accelerating embodied AI development and robotics.
RANK_REASON Release of a new multimodal world model (Atlas) by a prominent AI lab (World Labs) and a leading performance on an embodied world model benchmark (WorldArena 2.0) by a competitor (InSpatio). [lever_c_demoted from frontier_release: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
