XPENG Invited to CVPR for the Third Time, Showcasing China's Advances in Physical AI to the World
XPENG is showcasing its advancements in physical AI at CVPR 2026, marking its third attendance at the prestigious conference. Dr. Xianming Liu, Head of General Intelligence Center at XPENG, shared insights on the company's VLA2.0 foundation model, which has now entered mass production and achieved over 50% assisted driving mileage share in China. XPENG also unveiled its technical blueprint for a world model designed for deliberative reasoning, controllable generation, and long-horizon forecasting, complementing VLA2.0 by learning how the world changes after an action. The company is accelerating the large-scale deployment of its core Physical AI applications, including VLA2.0, Robotaxi, and Humanoid Robots, with the IRON humanoid robot progressing towards mass production by the end of 2026. These developments underscore XPENG's ambition to become a global leader in AI mobility and drive the Smart EV revolution.
XPENG's consistent presence and technical disclosures at CVPR highlight China's growing prowess in AI, particularly in the domain of physical AI and autonomous systems. The mass production of VLA2.0 and its rapid adoption in assisted driving in China signals a significant leap from research to commercial deployment, setting a new benchmark for the industry. This demonstrates not only XPENG's engineering capabilities but also the maturity of the Chinese market in embracing advanced AI-driven mobility solutions. The emphasis on a "world model" capable of complex reasoning and prediction indicates a strategic push towards more robust and generalizable AI, moving beyond reactive systems to truly intelligent autonomy. This approach, combining VLA2.0's action learning with the world model's environmental understanding, positions XPENG at the forefront of developing comprehensive physical-world foundation models. The rapid scaling of GPU training efficiency further underscores the intense investment and progress in AI infrastructure within China. The planned deployment of humanoid robots as shopping guides by early 2027 also points to a broader vision of integrating AI across various sectors, extending beyond automotive to retail and service industries. This diversification of AI application areas by a leading Chinese tech company reflects a national strategy to leverage AI for economic transformation and global competitiveness. The publication of academic papers alongside commercial deployment also signifies a commitment to open science and contributing to the global AI research community, while simultaneously building proprietary advantages. This blend of academic rigor and commercial ambition is a hallmark of China's innovation ecosystem.
