Large models accelerate their rollout in vehicles: large-scale deployment of intelligent connected vehicles still needs to clear bottlenecks
At the 2026 World Intelligent Connected Vehicles Conference, automakers and various players across the industry chain focused on large-scale deployment after large models are integrated into vehicles, discussing bottlenecks such as safety validation, cost control and division of labor across the industry. NIO said its intelligent assisted driving system based on world models and reinforcement learning has been deployed in more than 950,000 vehicles, and Changan Automobile expects L3 autonomous driving to enter mass production in 2027 and L4 commercial operations in 2028. Horizon Robotics' Yu Kai said urban assisted driving has entered a usable stage, but sustained efforts are still needed before it matures.
Body · AI translation
Related topics top start Related topics top end Futures promotion begin Futures promotion end Miaopai begin Miaopai end Video player start Video player end Market chart begin Market chart end
On October 21, the 2026 World Intelligent Connected Vehicles Conference opened in Beijing. As automotive large models and autonomous driving move out of the laboratory, the industry focus has shifted from breakthroughs in individual technologies to large-scale deployment that is safe, controllable, and coordinated in an orderly manner. From algorithm iteration to mass-production validation, from risk management to cross-domain collaboration, how to clear the bottlenecks in commercial deployment became the most closely watched topic at the event.
● Yin Haonan, Gong Mengze
Strengthening integrated application of artificial intelligence
As large models accelerate their adoption in vehicles, automakers' technology strategies are extending from developing individual functions to intelligent coordination across the whole vehicle.
Lei Jun, chairman and CEO of Xiaomi Group, believes that improving the intelligent experience of the whole vehicle requires relying on a unified and powerful foundation large model to drive coordination among different functions.
Technology application is already backed by fleet scale. Li Bin, chairman and CEO of NIO, said that its intelligent assisted driving system based on world models and reinforcement learning has been deployed in more than 950,000 vehicles. Looking ahead, Zhao Fei, general manager of China Changan Automobile Group, expects L3 autonomous driving to reach mass production in 2027 and L4 to achieve commercial operation in 2028.
Facing application needs at different levels and in different scenarios, how to balance performance and cost is also a question that mass-production development must answer. Zhong Zhihua, a member of the Presidium of the Chinese Academy of Engineering, proposed that promoting AI-enabled transformation and development of automobiles should be application-oriented and people-centered, supporting diverse product designs through shared common architectures, resources, and capabilities; and that control solutions should be selected for specific scenarios to balance safety, reliability, and cost-effectiveness.
Moving further from single-vehicle applications to the transportation system, coordinated development has also drawn attention. Wan Gang, honorary chairman of the China Association for Science and Technology, proposed adhering to the combination of single-vehicle intelligence and connected coordination, the combination of scenario-driven development and efficiency upgrading, and the combination of implementing phased results and exploring long-term goals. Li Yang, vice minister of transport, said that the Ministry of Transport will thoroughly implement the "AI + Transportation" initiative, strengthen research on key technologies, and use scenario applications to promote the integrated development of intelligent connected vehicles and transportation.
Prioritizing safety and quality
As the scope of intelligent driving applications expands, how to establish a quantifiable and traceable safety validation system has become another key focus of discussion at the conference.
Yu Kai, founder and CEO of Horizon Robotics, believes that urban assisted driving has now entered a usable stage, but sustained efforts are still needed before maturity. Each order-of-magnitude increase in the interval between takeovers may bring new technical challenges.
Regarding application results, Li Bin proposed that data from third-party institutions such as insurers can reflect changes in accident probability and accident rates, allowing users to perceive safety improvements and testing commercial value through users' willingness to pay. He Xiaopeng, chairman and CEO of XPeng Group, also regards safety, scenario coverage, and user payment as important evaluation dimensions, focusing on whether technological progress can be translated into actual experience and returns.
Safety management within companies is also being strengthened. Lei Jun said Xiaomi has established an automotive safety research center to independently investigate and analyze problems and then feed the results back to the R&D process for improvement, while also setting up institutions such as an all-domain safety department. Koersten, president and CEO of BMW Group Greater China, said that while pursuing innovation and speed, quality and safety must not be sacrificed.
Regarding safety requirements that the industry should jointly follow, Liu Yigong, general manager of China FAW, proposed that while promoting technological innovation and strengthening industrial collaboration, the industry must jointly safeguard the safety bottom line.
Feng Changjun, general manager of Dongfeng Motor Group, suggested strictly implementing automotive safety standards and testing and evaluation requirements, and promoting alignment and mutual recognition between Chinese standards and international standards.
Xin Guobin, vice minister of industry and information technology, said that the Ministry of Industry and Information Technology will ensure the promotion and implementation of mandatory national standards related to autonomous driving, accelerate the development of key standards such as operation monitoring and data security, thoroughly implement the special action for improving automobile quality, and build a sound policy environment for industrial development.
Deepening industrial division of labor and cooperation
The cost pressure of large-scale mass production is forcing the industry to move away from a go-it-alone model and use industrial division of labor and collaboration to resolve the resource waste of duplicated R&D.
Li Shufu, chairman of Geely Holding Group, said that in September this year Geely and NIO reached a comprehensive strategic cooperation in the battery swapping field, increasing energy replenishment options for users through open sharing, unified standards, and resource coordination. He believes that intelligent technologies should likewise be jointly built through collaboration to reduce duplicated investment and promote sustainable development.
In autonomous driving development, cooperation already involves underlying architectures and mass-production delivery. Cao Xudong, founder and CEO of Momenta, said that when developing jointly with automakers, a single L4 architecture can support L3 and L4 requirements, reducing the investment of developing two separate solutions. Yu Kai said Horizon Robotics focuses on computing chips and basic software, and currently more than 95% of vehicle model projects are delivered through partners.
Regarding industrial competition, Lei Jun said that in the second half of intelligent connected vehicles, competition involves not only vehicle intelligence levels but also the industrial ecosystem. China has a complete industrial chain, an active atmosphere for AI innovation, and a huge market, and should promote intelligent ecosystem connectivity and collaboration.
The scope of cooperation is also extending overseas. Yin Tongyue, chairman of Chery Holding Group, drawing on export experience, said that internationalization requires integrating into local industrial systems, supply chains, technology chains, and value chains, and meeting market access requirements with quality systems and technological strength. Xin Guobin proposed supporting Chinese and foreign enterprises in deepening R&D, investment, and industrialization cooperation to jointly create high-quality products.
At the conference, seven national technical standardization committees in the fields of automobiles, intelligent transportation, communications, traffic management, geographic information, smart cities, and road transport signed a framework agreement on Internet of Vehicles standards cooperation to jointly advance research, formulation, revision, and implementation of technical standards. Guo Shougang, director of the First Department of Equipment Industry of the Ministry of Industry and Information Technology, said that the high-quality development of intelligent connected vehicles cannot be achieved without cross-industry collaboration, and that establishing a cross-committee cooperation mechanism in the standardization field is of great significance. news_keyword_pub,stock,sz000625 iframe below main text iframe below main text General statement General statement QR code at end start QR code at end start
QR code at end start QR code at end start Editor name and work code
Editor name and work code