Nio Driver-Assist Software Upgrade Reaches 700,000 Vehicles With Advanced World Model Technology

Nio has begun rolling out the latest version of its Nio World Model (NWM) driver-assist software to vehicles, with more than 700,000 cars eligible for the upgrade. The company stated that customers who purchased their vehicles as far back as four years ago will also gain access to the newest technology. The update is designed to enhance the driver-assist training framework and deliver a more natural, human-like driving experience while extending advanced software capabilities across a large installed vehicle base.

Nio Achieves Cross-Platform Driver-Assist Software Deployment

Nio said it has become the first automaker to achieve synchronized deployment of co-developed driver-assist software across both general-purpose chip platforms and internally developed chip platforms. As a result, vehicles powered by four Nvidia Orin-X processors and models equipped with the company's Shenji NX9031 chip can receive identical software releases simultaneously. According to the company, this capability stems from years of full-stack in-house development efforts that began in 2020, supported by long-term investments in hardware architecture and software engineering capabilities.

In-House Engineering Supports Large-Scale AI Deployment

At the hardware level, Nio anticipated memory bandwidth limitations that could affect the deployment of large AI models inside vehicles and planned its computing resources accordingly. On the software side, the company chose not to rely on standard deployment toolchains supplied by chip vendors when using Orin-X platforms. Instead, it developed its own deployment framework and compiler technology. Nio stated that its internally developed AI compiler reduced the development cycle for new model operators from one to two weeks down to one to two days while increasing inference performance by 20 percent.

Nio Introduces Enhanced Three-Layer AI Training Architecture

The latest release represents another significant software milestone following the version introduced earlier in the year. In that earlier rollout, Nio became the first automaker in China to apply complete closed-loop reinforcement learning to driver-assist development. The new software expands this foundation through a three-layer architecture consisting of world modeling, supervised fine-tuning, and closed-loop reinforcement learning. By incorporating supervised fine-tuning based on high-quality behavioral datasets, the system is trained to replicate more human-like driving characteristics while maintaining compliance and operational consistency.

Key Features of the Latest NWM Update

• Three-layer AI training framework combining world modeling, supervised fine-tuning, and reinforcement learning.
• Direct output of steering, acceleration, and braking control signals.
• Improved route-selection accuracy without dependence on high-definition maps.
• Real-time recognition of tidal lanes and variable-lane overhead signs.
• Enhanced balance between false-braking reduction and risk intervention performance.

End-to-End Vehicle Control Improves Responsiveness

Nio stated that the new version is the first driver-assist system in China capable of directly outputting steering wheel and acceleration or braking pedal signals instead of generating sampled driving trajectories. The company believes this end-to-end kinematic modeling approach creates a shorter decision-making path and reduces latency. As a result, vehicle control becomes smoother and more precise, helping improve both comfort and operational accuracy during assisted driving scenarios.

NWM Driving Capability Highlights

The company reported that route-selection accuracy has reached an industry-leading level even without relying on high-definition mapping or enhanced navigation maps. In addition, Nio's NWM platform became the first in-house developed automaker system to achieve real-time recognition and interpretation of tidal lanes and variable-lane overhead traffic signs. Nio also emphasized that the software delivers strong performance in two traditionally conflicting areas: minimizing false braking events while maintaining effective risk intervention capabilities.

User Adoption Improvements Since Previous Release

Since the rollout of the earlier software version in January, Nio reported that both user engagement duration and utilization rates of its driver-assist functions have doubled. The company views these improvements as evidence that continuous software enhancements and more natural driving behavior are encouraging greater adoption of assisted driving technologies among vehicle owners.

Nio and Onvo Software Rollout Overview

Onvo Benefits From Shared Software Foundation

Meanwhile, Onvo, the mass-market sub-brand of Nio, has also started deploying the Coconut+ 3.0.2 software version using the same technical framework. Because the underlying software stack is developed internally, Nio said it can migrate mainline features to Onvo models running on the Shenji NX9031 platform in as little as two weeks. The company also plans to deliver a major software upgrade covering both smart cockpit and driver-assist functions to all Orin-X versions of Onvo vehicles before the end of July.

Focus on Higher Driver-Assist Usage

Nio stated that the objective of the latest rollout is to combine a forward-looking technology architecture with internally developed systems engineering expertise. By improving driving performance, reducing latency, enhancing human-like behavior, and expanding advanced functionality across multiple hardware platforms, the company aims to further increase the adoption and everyday usage of driver-assist technologies among its users.

Frequently Asked Questions

What is the latest Nio driver-assist software update?
The latest Nio driver-assist software update is a new version of the Nio World Model platform deployed to more than 700,000 vehicles. It introduces a three-layer AI training architecture that combines world modeling, supervised fine-tuning, and closed-loop reinforcement learning. The software also supports end-to-end vehicle control by directly generating steering, acceleration, and braking commands. These enhancements are intended to improve driving smoothness, route-selection accuracy, and overall driver-assist performance across compatible Nio vehicles.

How does Nio's new World Model improve assisted driving?
Nio's new World Model improves assisted driving by using advanced AI training methods and direct vehicle-control outputs to reduce latency and increase precision. The platform can learn from high-quality behavioral data to emulate more human-like driving behavior while maintaining safety and compliance. Additional capabilities include real-time recognition of tidal lanes and variable-lane signs, improved route planning without reliance on high-definition maps, and better balancing of false-braking prevention with effective risk intervention during driving scenarios.

Can older Nio vehicles receive the software upgrade?
Yes, Nio stated that even customers who purchased their vehicles up to four years ago can receive the latest software upgrade if their vehicles are compatible with the supported platforms. The company designed the rollout to benefit a large installed base while maintaining synchronized software releases across Nvidia Orin-X-powered vehicles and those equipped with the Shenji NX9031 chip. This approach enables existing owners to access many of the newest driver-assist capabilities without requiring a new vehicle purchase.