Chinese electric vehicle maker Nio has begun deploying a major update to its Nio World Model (NWM) driver-assistance software, with more than 700,000 vehicles eligible to receive the upgrade.
The company said the rollout extends to customers who purchased vehicles as far back as four years ago, allowing older models to access the latest generation of Nio’s driver-assistance technology.
Unified Rollout Across Multiple Chip Platforms
Nio said the latest software version is the first in the automotive industry to be simultaneously deployed across vehicles powered by both third-party and proprietary computing platforms.
The update is available for vehicles equipped with four Nvidia Orin-X processors as well as models powered by Nio’s in-house developed Shenji NX9031 chip.
According to the company, achieving synchronized software releases across different hardware architectures is the result of its full-stack software and hardware development strategy, which began in 2020.
Nio said it also invested early in computing resources and software infrastructure to support the deployment of increasingly complex AI models inside vehicles.
New AI Framework Introduced
The software update follows a major release introduced in January, when Nio said it became the first Chinese automaker to apply complete closed-loop reinforcement learning to driver-assistance development.
The latest version expands that approach with a three-layer AI training framework consisting of a world model, supervised fine-tuning, and closed-loop reinforcement learning.
Nio said the supervised fine-tuning component uses high-quality behavioral data to help the system learn driving patterns that are intended to more closely resemble human driving behavior while maintaining compliance with traffic rules.
Shift Toward End-to-End Vehicle Control
A key feature of the new software is its adoption of end-to-end vehicle control.
Rather than generating driving trajectories that are subsequently interpreted by vehicle systems, the software directly outputs steering, acceleration, and braking commands.
Nio said this approach reduces processing latency and improves the smoothness and precision of vehicle responses.
The company added that the updated system delivers improved route-selection performance even without relying on high-definition maps or enhanced navigation maps.
New Recognition Capabilities Added
The latest version also expands the system’s environmental understanding capabilities.
According to Nio, the software can now identify and interpret tidal traffic lanes and overhead variable-lane control signs in real time.
The company said the update improves the balance between minimizing unnecessary braking events and maintaining appropriate safety interventions, two metrics that can often conflict in driver-assistance systems.
In-House Development Accelerates Deployment
Nio said its internally developed AI compiler has significantly shortened software development cycles.
According to the company, development time for new model operators has been reduced from one to two weeks to as little as one to two days, while inference performance has improved by approximately 20%.
The company attributed these gains to its decision to develop its own deployment framework and compiler rather than relying exclusively on software toolchains provided by chip suppliers.
Onvo Models Also Receive Updates
The rollout extends beyond the Nio brand.
Onvo, Nio’s mass-market vehicle division, has begun deploying its Coconut+ 3.0.2 software version, which uses the same underlying technology framework.
Nio said its in-house software architecture allows new features to be transferred to Onvo models equipped with the Shenji NX9031 platform in as little as two weeks.
The company also plans to release a broader software upgrade for Onvo vehicles equipped with Nvidia Orin-X hardware before the end of July.
Usage of Driver-Assist Features Increasing
Nio said adoption of its driver-assistance functions has increased significantly since the January software release.
According to the company, both usage rates and total usage time of its driver-assist features have doubled during that period.
The latest update is intended to further increase adoption by improving system performance while leveraging Nio’s internally developed software and chip technologies across its vehicle lineup.
