Pony.ai has unveiled PonyWorld 2.0, a major upgrade to its proprietary world model designed to accelerate the development and deployment of autonomous driving systems.
"PonyWorld 2.0 is an important step toward a more self-improving approach to autonomous driving development," said Dr Tiancheng Lou, founder and CTO of Pony.ai. "As AI systems become more capable, they can play a larger role not only in learning to drive, but also in guiding their own improvement — making L4 development more scalable over time."
PonyWorld 2.0
The new system introduces self-diagnosis capabilities, allowing AI models to identify their own weaknesses, guide targeted data collection, and prioritise training on complex driving scenarios.
Unlike traditional simulation tools, PonyWorld has evolved into a full reinforcement learning system that supports both cloud-based training and real-world vehicle deployment. According to Pony.ai, improving its "Virtual Driver" now depends on enhancing the accuracy and realism of the world model used to train it.
The upgraded platform features a structured intention layer, enabling the system to analyse its decisions, compare outcomes, and generate targeted data collection tasks. Human teams then gather relevant real-world data to refine the model, to create a continuous feedback loop.
PonyWorld 2.0 has already been deployed across the company's Level 4 fleet, improving safety, ride comfort, and traffic efficiency. Following successful robotaxi operations in two major Chinese cities, Pony.ai is expanding globally, aiming to deploy over 3,000 vehicles across 20 cities by year's end.
