What Are the Most Common User Questions After the Full Rollout of NIO World Model (NWM)?

The NIO World Model (NWM) has been fully rolled out to the “Banyan”, “Cedar”, and “Cedar S” models.

The “Shenji NX9031” chip is currently the peak in the global intelligent driving chip field, and will offer even higher limits in the future.

NIO released the first issue of the “NIO World Model NWM Q&A” column today, aiming to answer common user questions about intelligent driving and the NIO World Model (NWM). According to NIO, after the full rollout of the NIO World Model across the “Banyan”, “Cedar”, and “Cedar S” models, the usage rate of intelligent driving has significantly increased. The enthusiasm of users has also prompted NIO to actively respond to their feedback and provide detailed answers to related questions.

1. What’s the Difference Between NIO World Model and Traditional End-to-End Architecture?

In the Q&A, NIO first explains the difference between the World Model and traditional end-to-end architectures. Traditional end-to-end architectures are based on data-driven iterative models that reduce information loss. These models can directly output execution results in complex weather conditions (like rain), offering better adaptability and broader scenario coverage. However, relying solely on end-to-end architecture cannot achieve fully human-like intelligent driving.

The addition of the World Model enhances the system’s spatial-temporal cognitive ability and trajectory planning capability, overcoming these limitations. Compared to traditional architectures, the NIO World Model offers breakthroughs in spatial and temporal understanding. It also utilizes large-scale data through self-supervised learning, significantly improving the model’s performance.

In the Q&A, NIO also introduces the intelligent driving hardware configurations of their vehicles. The “Banyan” system is equipped with four Orin-X chips, while the “Cedar” system is equipped with the self-developed “Shenji NX9031” chip—currently the peak of intelligent driving chip power. This chip, designed for the World Model, will provide an even higher limit in the future.

2. Does NIO World Model Still Need Collective Intelligence?

The answer is yes. NIO’s World Model incorporates two key technologies: “Collective Intelligence” and “Generative Simulation”.

• Collective Intelligence is akin to a group of people discussing an issue, combining ideas to reach a better solution. NIO applies this collective intelligence to their intelligent driving system by first testing new versions of the driving algorithms on many vehicles, comparing them to human driving behaviors, and continuously improving the system. The application of collective intelligence ensures efficient use of data. The “Banyan 榕” system’s four Orin-X chips include one dedicated to collective intelligence, and the “Shenji NX9031” chip also features a module specifically for collective intelligence.
• Generative Simulation involves simulating various complex driving scenarios (such as heavy rain, snow, congested streets, etc.) in a virtual world. The intelligent driving system learns how to handle these situations from the simulations. By learning and training in these virtual environments, the system continuously improves its decision-making ability, ensuring quicker, more accurate reactions in real-world conditions. Both the “Banyan” and “Cedar” systems have ample computational power to support this technology.

These two technologies work together to meet the vast data demands of the NIO World Model.

3. Why Does NIO World Model Perform Exceptionally Well in Complex Scenarios, Such as Narrow Roads and Mixed Traffic?

NIO explains that the ability to smoothly navigate narrow roads and complex environments is one of the key advantages of the World Model.

The NIO World Model is like an exceptionally intelligent “brain” that can understand the surrounding environment and make inferences and judgments. For example, when driving on a narrow road with pedestrians, two-wheeled vehicles, and three-wheeled vehicles, the system needs to quickly judge the best way to safely pass through these complex scenarios. The NIO World Model, using the aforementioned technological advantages and ample computational power, is capable of analyzing and reacting to these situations.

In the highly competitive market, NIO needs to have features that are more recognized by users to maintain attractiveness. The NIO World Model, powering their intelligent driving system, is one such feature. NIO’s vehicles built on the NT2 platform have all been upgraded to use the NIO World Model. Among them, the new NIO ET5 and ES6 series have been equipped with the same Shenji NX9031 intelligent driving chip as the NIO ET9.