China has announced the first group of automakers cleared to pilot L3-level autonomous driving, marking another step in the gradual move toward fully self-driving vehicles.
The announcement follows the release of the “Work Plan for Steady Growth of the Automobile Industry (2025–2026)” earlier this year.
According to China’s “Classification of Automobile Driving Automation,” L3 technology falls under conditional autonomous driving.
It allows drivers to take their hands off the wheel in specific scenarios while remaining ready to intervene at any time.
The Ministry of Industry and Information Technology (MIIT), together with the Ministry of Public Security and other departments, have recently published the first batch of pilot approvals covering nine automaker consortiums across passenger cars, buses, and trucks.
Changan leads the group with 17 testing licenses, the highest among participants. Its approved models include the Deepal SL03 and the Avatr 11, 12, 07, and 06, all equipped with Huawei’s ADS 4.0 system.
Avatr is jointly owned by Changan, CATL and Huawei and showcased last month — at the IAA Mobility Show in Germany — its new concept model, the ‘Vision Xpectra’ sedan.
Seres, also known by its Aito brand, was approved for its Wenjie M9, M8, and M7 models, which use Huawei’s ADS 4.0 software and LiDAR sensors.
BYD was listed for vehicles featuring its in-house “God’s Eye” system, announced in early 2025. The system has already completed highway Navigate on Autopilot (NOA) testing.
God’s Eye is offered in three configurations.
The top-end God’s Eye A, or DiPilot 600, employs three LiDAR sensors and is mainly used by BYD’s ultra-luxury Yangwang brand. God’s Eye B, or DiPilot 300, integrates one LiDAR and equips Denza and certain BYD premium models.
God’s Eye C, or DiPilot 100, is a camera-only version without LiDAR used in more affordable vehicles, including the entry level model BYD Seagull — known as Dolphin Surf in Europe.
The system includes features such as automated parking, adaptive cruise control, and advanced braking assistance.
Nio received approval for an unreleased model equipped with its self-developed NWD (Nio World Model) autonomous-driving system, which integrates Nio’s in-house Shenji NX9031 chip.
The Shanghai-based automaker began developing chips in 2021 and has since launched two proprietary designs: the Yangjian chip for LiDAR control and the Shenji NX9031 chip for autonomous driving.
The Shenji chip, unveiled in late 2023, was first integrated into Nio’s flagship ET9 sedan and later deployed in updated versions of the ES6, EC6, ET5, and ET5 Touring, which launched in mid-May.
Founder and CEO William Li said recently that Nio spent over $300 million on Nvidia Orin-X chips last year. He added that shifting to its own Shenji processors could save about 10,000 yuan ($1,400) per vehicle.
The Shenji chip marks Nio’s move away from reliance on Nvidia hardware, though its Onvo-branded vehicles continue to use Nvidia chips.
As reported earlier this week, Nio’s head of intelligent driving product and experience, Harry Wong, resigned from the EV maker.
The executive was one of the key team members behind Nio’s assisted driving software, the Nio World Model.
In June, Chinese outlet LatePost reported that Nio was seeking strategic investors for its chip business, a claim the company denied.
The automaker had earlier established a new legal entity, Anhui Shenji Technology Co., Ltd., with a registered capital of 10 million yuan ($1.4 million), covering chip design and sales as part of its broader push into autonomous-driving technologies.
Notably absent from the L3 pilot list are Tesla, XPeng, and both Beijing-based brands Li Auto and Xiaomi.
In the US, Tesla is releasing the V14 of its Full Self Driving (Supervised) software this Monday, October 6.
Elon Musk has previously stated that the new version will be the second biggest update ever for Tesla’s assisted driving technology.
XPeng’s founder and CEO He Xiaopeng said in February that the company planned to roll out L3-level autonomous-driving software by year-end, emphasizing the need for a several-fold increase in computing power and storage capacity.
