Li Auto is on track to offer its new level 3 automated driving system within the next 12 months, which would enable hands-free, eyes-off driving under certain conditions, as it transitions towards “end-to-end neural network” architecture, according to chief executive Li Xiang.

Why it matters: The comments mark an increase in competition between Tesla and Chinese automakers over the adoption of autonomous driving. The US automaker is reportedly looking to roll out the latest version of its “Full Self-driving-Driving” software in China, its second-largest market.

  • The news also comes as Li Auto started calling back some former staff late last month having previously dismissed roughly 30% of its 1,100-person Autonomous Driving Group in a setback to the company’s launch of its first battery electric model, the Mega
  • Most employees rehired by the electric vehicle maker work on the NOA (Navigation on ADAS — advanced driver assistance system) function for urban driving scenarios. Vice President Lang Xianpeng on June 7 promised surviving employees from his team that the layoff was now over, sources familiar with the matter told the 21st Century Business Herald (in Chinese).

READ MORE: China opens door wider to Tesla as local giants disrupt the EV sector with AI-defined vehicles

Details: Li Auto will roll out a new version of its NOA software in the third quarter of 2024. This version still requires the driver to be in full control but will not rely on a high-definition map. The company announced that it will launch its city NOA function to users in 100 major Chinese cities in November.

  • This will be followed by a wide release of its Level 3 autonomous driving system later this year, or in early 2025, Li told the audience at the China Auto Chongqing Summit on June 8, according to short videos posted on the social media platform WeChat.
  • Li said the company has been working on an end-to-end artificial intelligence approach since late 2023. The approach is using video clips from real driving situations as well as machine learning algorithms to handle the vehicles’ control. It is also exploring Visual-Language Models (VLMs), which could better learn patterns and correlate between visual data and natural language than existing offerings.
  • Li Auto’s L3 conditional autonomous system will use both approaches, Li said, allowing the human driver not to monitor the vehicle constantly but to be ready to take control, according to the definitions set by the Society of Automotive Engineers (SAE). It could further evolve to a level 4 highly autonomous driving system over the next three years, Li added, without elaborating further.

Context: Automakers in China, including Li Auto, have either been applying for mapping licenses or working with map services in order to collect large amounts of highly detailed geographical data, which helps autonomous vehicles (AVs) make better decisions when performing driving tasks. However, many of them, such as Huawei and Xpeng Motors, are now looking for alternatives due to cost and complexity reasons.

  • It is unknown whether Li Auto will take legal responsibility for its upcoming level 3 technology. Last year, Mercedes-Benz became the first automaker allowed to sell cars with its L3 Drive Pilot system in California and Nevada, Reuters reported. China is pushing for similar regulatory initiatives with assistance from automakers including BYD, SAIC, and NIO.
  • Li Auto has been working on layoffs since May to cut 18% of the company’s workforce, which stood at roughly 5,600 as of last year, while that percentage for the Autonomous Driving Group was more than 30%, according to the 21st Century report. President Ma Donghui told investors in November the company planned to have 2,000 employees for AVs by the end of 2024. 

READ MORE: Li Auto accelerates assisted driving tech competition amid launch of first battery EV

Jill Shen is Shanghai-based technology reporter. She covers Chinese mobility, autonomous vehicles, and electric cars. Connect with her via e-mail: jill.shen@technode.com or Twitter: @jill_shen_sh