On April 25, 2026, Momenta held a thematic session titled “Momenta R7: The Prologue of Physical AI” at the Beijing International Auto Show. Four co-founders jointly announced the mass production debut of Momenta R7’s reinforcement learning-based world model. This technology marks the evolution of intelligent driving from merely “seeing the world” to truly “understanding the world,” signifying that Physical AI has officially entered the era of large-scale mass production. Momenta has already delivered more than 70 mass-produced vehicle models, with over 200 models awarded定点 (定点 refers to official selection or nomination for integration), and its fleet of mass-produced vehicles has surpassed 800,000 units—capable of delivering 100,000 units in as little as 40 days. At this year’s Beijing Auto Show, over 20 brands and more than 60 vehicle models feature Momenta’s solutions, including newly launched models from Mercedes-Benz, Audi, and BMW. Momenta’s system is now deployed in more than ten countries and regions, including the UK, Norway, Singapore, Australia, and New Zealand. Momenta defines the core of Physical AI as the integration of world models and reinforcement learning. The world model enables the system to understand object properties, causal relationships of motion, and potential interactions by predicting future states and interaction logic within the physical world. Reinforcement learning, meanwhile, allows autonomous optimization through real-world environmental feedback. Together, these two components form the twin pillars of Physical AI. The world model of Momenta R7 consists of three layers: - The first layer involves pre-training, leveraging massive real-world driving data to distill physical laws, common sense, and causal relationships. - The second layer employs simulation for closed-loop autonomous driving reasoning, evaluating performance in long-tail scenarios. - The third layer conducts reinforcement learning in a highly realistic virtual environment, enabling the system to autonomously learn optimal decision-making strategies through repeated trial and error in complex interactive situations. In extreme long-tail scenarios—such as cargo falling from a vehicle ahead—Momenta R7 can anticipate object trajectories and plan appropriate responses based on its understanding of physical laws, rather than relying on pre-programmed rules or memorized patterns. This capability aims to enable the system to outperform human drivers in rare and challenging situations. Additionally, Momenta continues to expand its Robotaxi ecosystem, having established partnerships with mobility service providers and OEMs including Shouqi Mobility, IM Motors, Uber, Mercedes-Benz, AutoGo, Lumo, and Grab to accelerate the global commercial deployment of autonomous driving services.