XPeng and Academic Partners Release GroundingPI Open-Source Vision Model for Physical AI
Researchers from Chinese automaker XPeng and academic partner institutions have introduced GroundingPI, an open-source visual perception foundation model engineered for physical AI and spatial tasks. The 4-billion-parameter architecture targets enhanced spatial comprehension and visual grounding across dozens of computer-vision benchmarks. The project aims to bolster the environmental perception capabilities underpinning advanced driver-assist and autonomous driving software.