On April 25, 2026, Momenta held a themed event titled "Momenta R7: The Physical AI Era Begins" at the Beijing International Auto Show. Marking the company's tenth anniversary, its four co-founders took the stage together for an in-depth dialogue on Momenta's technology philosophy, product capabilities, and global commercialization, and officially announced that the Momenta R7 World Model has entered mass production for the first time – signaling a leap in intelligent driving from "seeing the world" to "understanding the world," as Physical AI formally moves from a technology concept to large-scale mass production.
Understanding the Essence of Intelligence: Achieving General Intelligence with Physical AI
Momenta Co-founder and CEO Xudong Cao offered an in-depth breakdown of the underlying logic and evolutionary path of Physical AI. Prediction is the core cornerstone of intelligent evolution: large language models, powered by Next Word Prediction, compress the common-sense knowledge of the digital world, giving AI the ability to understand text and natural language; world models, through World Model Prediction, predict the future states and interaction logic of the physical world, gaining the ability to understand the physical properties of objects, the causal relationships of motion, and the potential possibilities of interaction.
Self-evolution is the key driving force behind continuous iteration of intelligence. Within its environment, and guided by its objectives, the AI continuously receives feedback, iterates, and self-optimizes. This evolutionary logic is precisely the core of reinforcement learning.
Thus, world models and reinforcement learning together form the two core pillars of Physical AI.
World Models and Reinforcement Learning: Building the Cornerstone of Physical AI
Momenta Co-founder and SVP of R&D Yan Xia provided an in-depth breakdown of the technical architecture of the Momenta R7 World Model.
Xia noted that the core of Physical AI lies in a deep cognition of the fundamental physical laws of the world. Momenta's world model is structured in three layers.
The first layer, World Model Pre-training. Through pre-training on massive real-world data, physical laws, common sense, and causal relationships are compressed into the model, giving the system a foundational understanding of the physical world.
The second layer, World Model Simulation. The world model is applied to closed-loop simulation for autonomous driving, enabling the system to extrapolate how the world will evolve as its own behavior changes and, drawing on efficient scenario extrapolation, to evaluate performance in long-tail scenarios.
The third layer, Reinforcement Learning within the World Model. Building on the first two layers, a highly realistic virtual training ground is created for reinforcement learning, allowing the system to explore and iterate repeatedly in a near-real environment.
It is precisely in this virtual training ground that reinforcement learning can fully realize its strengths. The system moves from mere "Imitation" learning to "Imagination" – undergoing tens of millions of extrapolations in the virtual world and autonomously mastering the ability to make optimal decisions in complex interactions. The one-in-ten-thousand extreme scenarios of the real world have already been experienced and fully absorbed, time and again, in the virtual training ground. The goal of this mechanism is to enable the model to surpass the level of human performance in rare extreme scenarios – and this is where the true value of Physical AI lies.
Exceedingly Rare Extreme Scenarios, Ultimate Safe-and-Reassuring Protection
Momenta Co-founder and SVP of R&D Gang Sun emphasized that true Physical AI is not only about letting vehicles operate smoothly in the vast majority of everyday scenarios, but also about continuing to provide users with ultimate, safe-and-reassuring protection in the one-in-ten-thousand rare extreme scenarios.
At the event, Gang Sun used several sets of real-world demonstrations to vividly showcase the Momenta R7 World Model's outstanding performance in extreme situations:
Predicting a suddenly falling apple
Acting in a split second
Navigating tight spaces with ease
Planning ahead like a rational operator
The long-tail extreme situations that traditional end-to-end algorithms struggle to anticipate and cover are the true proving ground for the value of Physical AI. The Momenta R7 World Model's composed handling of these situations confirms its core capability: truly understanding the laws of motion and the interaction logic of the physical world, rather than relying on scenario memory and rule-matching. This is precisely what enables the Momenta R7 World Model to provide users with comprehensive, reassuring protection in one-in-ten-thousand extreme scenarios.
China's Intelligent-Driving Flywheel Speed: A Trusted Partner of Global Leaders
Momenta Co-founder and SVP of Marketing & Business Development Huan Sun shared the latest progress in the company's global commercialization.
The data speaks for itself. Over the past year, the number of mass-production vehicles equipped with Momenta's system has jumped from nearly 300,000 to more than 800,000 – an increase of over 160%. Mass-production speed has likewise made a qualitative leap: in 2022, the first 100,000 units took 24 months; today, 100,000 deliveries can be completed in as little as under 40 days. The flywheel effect is accelerating.
To date, Momenta has successfully delivered more than 70 mass-production models, with cumulative design wins (model nominations) exceeding 200. At this year's Beijing International Auto Show, more than 20 brands and over 60 models featured Momenta's intelligent assisted driving solutions – spanning a diverse matrix of domestic brands and Chinese-foreign joint-venture brands – the most powerful testament to its standing as a trusted partner of global leaders.
On the global front, Momenta's solutions have been deployed in more than 10 countries and regions, including the United Kingdom, Norway, Singapore, and Australia and New Zealand. At the same time, Momenta's global Robotaxi ecosystem continues to expand, having reached strategic partnerships with multiple ride-hailing companies and OEMs to jointly advance the large-scale rollout of commercial autonomous driving services and provide users worldwide with a safe, convenient, and efficient mobility experience.
Our Founding Aspiration Is Our Mission: Better AI, Better Life.
Reflecting on the company's founding aspiration, Xudong Cao recalled how, ten years ago in Silicon Valley, he came across a street named after Fairchild Semiconductor – "Fairchild Drive" – the birthplace of the global semiconductor industry and the street that ultimately shaped the legendary character of Silicon Valley. "In that moment, I felt an immediate spark; and today, we hope to join hands with all of China's AI companies to together write our own Silicon Valley story."
Ten years is both a starting point and a proof. Momenta's mission and vision remain firm: to save one million lives in ten years, to free 100% of driving time in ten years, and to double logistics and mobility efficiency in ten years. The launch of the Momenta R7 World Model opens the prologue of Physical AI and marks the starting point of the next decade. Momenta will join hands with global partners to keep pushing technological breakthroughs that bring intelligent driving to every user.
The opening chapter has begun, and we keep moving forward. Better AI, Better Life.
Back
