XPeng Motors recently announced that its second-generation Vision-Language-Action (VLA) physical AI large model has been adapted to work across both the Chinese and European markets with a single unified model. The brand is currently accelerating mass-production adaptation for the European market, with a target to gradually roll out deliveries to overseas users by 2027. XPeng has long maintained that autonomous driving is essentially a physical AI problem, and understanding the laws of the real physical world is the core foundation of its second-generation VLA model.

The second-generation VLA features native multimodal design, reconstructing the core of autonomous driving decision-making with integrated visual, audio and text perception. It boosts visual chain-of-thought reasoning efficiency by 32 times and can independently deduce optimal driving strategies. Critically, the model can autonomously generate training data to accelerate scenario generalization in overseas markets, eliminating the need to build local data sets from scratch. This will drastically shorten adaptation cycles for overseas markets and lay the technical foundation for XPeng’s global expansion.