Exclusive | World Model Startup Zhitianxia Secures Angel Funding, Positioning as 'China's Answer to World Labs'
Exclusive | World Model Startup Zhitianxia Secures Angel Funding, Positioning as 'China's Answer to World Labs'
Forging the path to world models via a generative 3DGS foundation model.
AI Tech Review has learned exclusively that Zhitianxia (Suzhou) Artificial Intelligence Technology Co., Ltd. — a spatial intelligence and world model venture — has closed its angel financing round.
The round was led solely by Shanghai Zhonghe Venture Capital, with Probe Capital acting as financial adviser. Proceeds are earmarked for team expansion in generative Gaussian foundation model R&D and for extending the overseas reach of China's Gaussian Splatting content community.
Zhitianxia specializes in 3D Gaussian Splatting (3DGS)-based 3D reconstruction and generation. The company rolled out a free 3DGS reconstruction and publishing service in early 2024, attracting tens of thousands of users and cultivating a community centered on technical exchange and content creation. It has also amassed a high-quality multi-view photo dataset. Daily 3DGS creation volume on its platform recently overtook that of the UK's SuperSplat, trailing only US-based World Labs.
Powered by petabytes of accumulated real-world 3D imagery, Zhitianxia is building a generative Gaussian foundation model engineered for continuous scale — designed for 3D spatial reasoning and completion. The effort aims to advance AI from text-based comprehension toward true 3D physical-world understanding, laying the low-level infrastructure for spatial intelligence.
3DGS has emerged as a priority technology domain for overseas tech firms in recent years. NVIDIA has embedded it into its Omniverse and Isaac Sim platforms; World Labs, founded by Fei-Fei Li, has likewise adopted it as its core data structure.
The technology represents scenes as a collection of optimizable 3D Gaussian ellipsoids, balancing photorealistic rendering quality, efficient real-time performance, and deep-learning differentiability. It has become a de facto spatial representation standard for embodied intelligence, AR/VR, digital twins, and autonomous-driving simulation.
Founder Dr. Song Kuan holds dual bachelor's degrees from Peking University and a Ph.D. from the University of Maryland, and has been recognized by talent programs in Beijing's Zhongguancun and Suzhou. In 2023, he charted a technical roadmap integrating 3DGS with multimodal and generative AI, and spearheaded the creation of China's largest 3DGS technical community.
Chief Scientist Wei Zeqiang has been affiliated with Dr. Song's team since 2019 and earned his Ph.D. from Beijing University of Posts and Telecommunications' School of Artificial Intelligence in 2025. Specializing in multi-view deep learning, he has led teams to multiple algorithm competition titles at CVPR, ECCV, and ICCV, and has authored multiple publications at premier AI conferences. (Leiphone)