In recent months, the Chinese automotive industry has witnessed a significant wave of partnerships with AI technology firm DeepSeek, igniting a trend of integrating large language models into vehiclesAccording to incomplete statistics compiled by Economic Observer, as of February 6, 2023, nearly 20 automotive enterprises and brands have announced deep collaborations with DeepSeek in the domains of smart cabins and AI operationsThis list includes prominent domestic car manufacturers such as Geely, Chery, Dongfeng, and Great Wall, as well as joint ventures like FAW-Volkswagen and Smart, not to mention new forces like Leap Motor, IM, and ZhiHu.

DeepSeek is a large language model developed by a domestic AI company, quickly gaining popularity due to its efficient and low-cost language comprehension and logical reasoning capabilitiesIts user base and traffic have surged, as the demand for smart vehicle technology continues to rise, making DeepSeek resemble a magic potion in the eyes of car manufacturers eager to enhance in-car smart interaction experiences.

As automotive smart technology becomes more prevalent, DeepSeek's role as a potential game-changer has drawn comparisons to the frenzy surrounding Baidu's release of the “Wenxin Yiyan” large model, which previously attracted numerous car manufacturers eager to incorporate it

Advertisements

However, the actual impact of the Wenxin Yiyan model on automotive intelligence remains mostly unexplored and undisclosed by the industry.

So, what results can be expected from this collective integration of DeepSeek by various car manufacturers? Can DeepSeek genuinely bring substantial improvements to vehicular interaction experiences and smart driving capabilities?

After DeepSeek's meteoric rise at the start of 2025, Geely was announced as the first automotive company to integrate this technologyOn February 6, Geely proclaimed that its self-developed Xingrui large model would undergo in-depth integration with DeepSeek-R1, utilizing the latter for distilled training of vehicle control and proactive interaction modelsSubsequently, other companies like Zeekr, Landau, Baojun, IM, Dongfeng, Leap Motor, Great Wall, Deep Blue, Smart, GAC Group, and Extreme Fox quickly followed suit with similar partnership announcements.

While most automotive players have integrated either self-developed or third-party intelligent interaction systems, many consumers still report shortcomings in achieving natural, smooth, and intelligent interactions

Advertisements

The integration of DeepSeek is viewed as a potential remedy to these consumer pain points. “Voice interaction can be seen in two layers; the first layer covers simple commands like adjusting the air conditioning or playing music, responding swiftlyThe second layer tackles more complex inquiries such as checking the weather or nearby restaurant recommendations, requiring intricate reasoning capabilities typically supported by a cloud-based large modelCurrently, DeepSeek primarily aims to enhance capabilities in this second layer,” noted Lu Wenliang, deputy director at the Industrial Tech Innovation Research Center at the Chinese Academy of Sciences.

Dongfeng Motor indicated that with DeepSeek's incorporation, voice interactions within vehicles would become more natural, demonstrating smarter scene comprehension and significantly improved iteration speed for functionalitiesBy employing multi-modal perception decision-making technology, the system can not only accurately discern driver intentions but also provide personalized recommendations based on both internal and external scenarios, thus evolving from a mere “transportation device” to a “smart mobility partner.”

The current trend of automotive manufacturers flocking to integrate DeepSeek underscores their commitment to advancing intelligent developmentOn February 5, He Xiaopeng, chairman of Xiaopeng Motors, highlighted in an opening address that “the Chinese tech company DeepSeek has amazed the global tech community, not only achieving product experience comparable to that offered by OpenAI’s large models but also pushing costs down to extremely low levels

Advertisements

In the next decade, AI will catalyze a massive transformation in automobiles and revolutionize the hardware and software industries worldwide.”

Presently, intelligent technology has become a key battleground in the development of new energy vehiclesResearch conducted by Kaiyuan Securities suggests that the integration of DeepSeek is expected to deliver superior cabin interaction experiences, resulting in unprecedented enhancements in smart cabin functionalities and potentially giving birth to entirely new application scenariosFurthermore, the model conserves computational power, facilitating easier local deployment within cabins for optimal experiences.

In fact, since the skyrocketing popularity of ChatGPT in 2022, automotive companies have progressively pushed for the incorporation of AI large models into their systemsSome new automotive players have even released their versions of AI large models, such as Li Auto’s MindGPT, Huawei’s Pangu Model 3.0, and Xiaopeng’s XGPT Lingxi ModelCurrently, brands such as “Weilai, Xiaopeng, and Lixiang,” and Hongmeng Zhi Xing have yet to announce their integration with DeepSeek, which may relate to their unique AI development strategies.

Following Baidu's launch of the Wenxin Yiyan large model in 2023, several automotive companies including Geely, Great Wall, Changan, Dongfeng Nissan, Leap Motor, IM, and Landau rushed to integrate this large model to elevate vehicle intelligence

Yet, details on the outcomes derived from this integration remain scarce from the manufacturersWill this new fervor surrounding DeepSeek yield different results?

There are perspectives suggesting a promising outlook for car manufacturers integrating DeepSeekHuafu Securities stated in their research report that as more automotive enterprises join DeepSeek's ranks, they can more swiftly dismantle technical barriers within the industry, reducing the application costs and thresholds of AI technology in smart automotive realms while delivering enhanced performance and service, ultimately improving the added value and competitiveness of smart vehicles.

Conversely, some experts assert that the partnerships between car manufacturers and DeepSeek mostly hinge on hypeLu Wenliang explained, “DeepSeek has made its source code open, and the cost of local deployment is relatively low and not challengingThe buzz around DeepSeek is set off to attract attention, but whether these integrations yield helpful results is yet to be seen.”

From the multitude of announcements regarding the integration with DeepSeek, it appears that this technology primarily focuses on voice control and training within the smart cabin rather than being closely associated with the more complex aspects of intelligent drivingWhether DeepSeek can significantly enhance intelligent driving capability thus emerges as another key factor in its scrutiny.

Kaiyuan Securities highlighted that through innovative architecture and algorithms, DeepSeek has greatly reduced memory usage and computational burdens, increasing the efficiency of computational resource utilization under limited processing capabilities and training costs

This could have profound implications for intelligent driving, expediting the pace of domestic autonomous driving development.

However, the prevailing sentiment in the industry is that large models are unlikely to bring immediate enhancements to intelligent driving capabilitiesGu Jianmin, Chief Technology Officer at Valeo China, stated that while DeepSeek supports the integration of AI in vehicles, the possibility of car manufacturers directly applying DeepSeek for intelligent driving is presently low due to differing information input sources and application scenariosMoreover, the model's stability presents significant challenges. “In my experience with DeepSeek, I have encountered several instances where it misjudges simple questions or decisionsWhen applied to intelligent driving, it needs a backup plan to ensure minimum safety standards; otherwise, it becomes irresponsible regarding driving safety,” he asserted.

Cui Dongshu, Secretary-General of the Joint Conference of Passenger Car Market Information at the China Automobile Circulation Association, also warned that while the integration of DeepSeek could foster improvements in intelligence levels, it is not a panaceaCar manufacturers must possess robust data and analytical capabilities because no matter how advanced a model may be, if the original data is tainted, the outcomes will never be accurate.

Ultimately, the extent to which DeepSeek can improve the driving experiences for consumers hinges on the automotive manufacturers’ ability to develop personalized and unique applications atop DeepSeek's foundation

Advertisements

Advertisements