Nio began rolling out the latest version of its Nio World Model (NWM) driver assistance software to vehicles in China on Wednesday, saying the update is designed to deliver a more human-like driving experience as the electric vehicle maker steps up development of advanced driver assistance technologies.
The update will initially be available to more than 460,000 vehicles equipped with Nio’s Banyan system, the company said. Models running the Cedar and Cedar S systems are expected to receive the software shortly.
See also: Nio Steps Up AI Push, Recruits Former Alibaba Expert to Strengthen Smart Driving Research
The Banyan system powers Nio vehicles built on its NT 2.0 platform and uses four Nvidia Orin X autonomous driving chips. The Cedar system is used in the Nio ET9 sedan and the third-generation ES8 sport utility vehicle, while the Cedar S system is installed in the 2025 model year ES6, EC6, ET5 and ET5 Touring, which are equipped with Nio’s in-house Shenji NX9031 autonomous driving chip.
Nio said this marks its second major update to the NWM software since late May 2025 and represents China’s first application of full closed-loop reinforcement learning in smart driving development. The updated system provides comprehensive modeling for both urban and highway piloted driving, delivering improvements in efficiency and driver reassurance, the company added.
See also: Nio Tests Vehicle-to-Grid Power Supply Using Battery Swap Stations in Jiangsu
The new version also enhances intelligent parking assistance and active safety features. In addition, it introduces pilot driving between battery swap stations in urban areas, which Nio described as the industry’s first smart driving function covering urban energy replenishment scenarios. The feature currently supports more than 2,000 second-generation and newer urban battery swap stations.
Ren Shaoqing, head of smart driving at Nio, unveiled NWM at the Nio IN 2024 Tech Day event in July 2024, describing it as “the world’s first smart driving world model.” According to the company, the model can simulate 216 potential driving scenarios within 100 milliseconds to identify optimal decisions.
See also: Nio’s Onvo Launches Lunar New Year Special Edition of L60 SUV
In the latest update, Nio said model training no longer relies on costly expert data, instead using large volumes of real-world human driving data combined with feedback mechanisms. “Through this approach, the system achieves both human-like reassurance and human-like efficiency,” the company said.
Nio said the update signals a new phase in its smart driving development and added that it plans to increase investment in cloud-based training computing power in 2026 to accelerate feature iteration under a new technical architecture and research and development framework.