XPENG Motors announced a significant expansion of its AI-driven mobility strategy with the upcoming global deployment of its VLA 2.0 intelligent driving system. The company confirmed that global deliveries will begin in 2027, following the commencement of international road testing. Volkswagen has been named the inaugural launch partner for the VLA 2.0 technology within the Chinese market.
Highlights
- XPENG VLA 2.0 features an end-to-end vision-to-action architecture that removes intermediate language layers to improve response speed and decision-making efficiency.
- The system is the first intelligent driving model in China with L4 potential and will be equipped on XPENG’s upcoming Robotaxi fleet.
- Global delivery of the VLA 2.0 system is scheduled for 2027, with public road testing for Robotaxis starting in China later this year.
- Beyond passenger vehicles, the VLA 2.0 architecture is designed for universal application across humanoid robots and modular flying vehicles.
Architectural Innovation and Technical Specifications
The VLA 2.0 system, which stands for Vision-Language-Action, represents a major shift from traditional modular autonomous driving stacks. While standard models use a sequential pipeline that translates visual perception into language-based reasoning, VLA 2.0 utilizes a unified AI foundation model. This architecture processes visual signals from the physical world directly into action commands.
The platform is powered by XPENG’s proprietary Turing AI chip. Each chip provides up to 2,250 TOPS of effective computing power, and vehicles utilize multiple chips to increase model inference efficiency. This hardware capability enables the deployment of large-scale AI models in production vehicles while reducing reliance on high-definition maps.
Performance and Full-Scenario Capabilities
The system is designed to provide a “driving-anywhere” experience, covering campus lanes, rural dirt roads, and non-navigable routes. In measured testing during Guangzhou’s evening rush hour, VLA 2.0 demonstrated a 23% improvement in driving efficiency compared to previous systems. Key performance features include:
- Human-like Reasoning: Interprets hand gestures from traffic police and initiates movement as traffic light countdowns end.
- Complex Navigation: Handles narrow alleys, residential complexes, and pothole avoidance with predictive awareness.
- Emergency Intervention: Detects driver incapacitation via in-cabin cameras and autonomously pulls the vehicle over to a safe stop.
- Standstill Start: Supports starting from a complete stop to provide a full-process assisted driving experience.
Smart Cockpit and Safety Features
The Surround Reality (SR) system connects with VLA 2.0 to provide real-time identification of pedestrians, electric scooters, and surrounding vehicle turn signals. This information is synchronized on the vehicle display to generate intuitive driving trajectories for passengers. Additionally, the in-cabin AI camera monitors the driver for fatigue or distraction, providing proactive alerts regarding mobile phone use or road attentiveness.
Global Roadmap and Robotaxi Trials
XPENG plans to begin trial operations for its VLA 2.0-equipped Robotaxis in China later in 2026. The company anticipates that full autonomy will arrive within the next one to three years as regulatory frameworks mature. While VLA 2.0 remains a supervised system for now, the underlying architecture serves as a foundation for broader embodied AI applications, including the company’s flying vehicle systems. More information is available at https://www.xpeng.com/.
Sign up for our popular weekly email to catch all the latest EV news!
