
Automakers are rushing to get involved in the Deepseek craze

Multiple car companies are collaborating with DeepSeek to enhance intelligent interaction experiences. Geely Auto announced that its Xingrui large model is deeply integrated with DeepSeek-R1, improving the in-car AI's understanding of user intentions. The ZEEKR Intelligent Tech team has also completed the integration of its self-developed Kr AI large model with DeepSeek-R1, enhancing predictive capabilities. Voyah is working with DeepSeek to optimize the intelligent experience of in-car systems, planning to launch its first mass-produced model integrated with DeepSeek through an OTA update on February 14. Other brands such as IM Motors and Dongfeng are also taking swift action, demonstrating DeepSeek's influence in the industry
Author | Chai Xuchen
Editor | Wang Xiaojun
The fire of DeepSeek has also reached the automotive circle.
On February 7th, Geely Auto was the first to announce that its self-developed Xingrui large model has completed deep integration with DeepSeek-R1. It is reported that Geely will use DeepSeek-R1 for distillation training of the Xingrui vehicle control Function Call large model, automotive proactive interaction endpoint large model, and more.
At that time, Geely's in-car AI will not only be able to accurately understand users' vague intentions and accurately call about 2,000 in-car interfaces, but also actively analyze users' potential needs based on in-car and external scenarios, providing users with proactive vehicle control, active dialogue, after-sales services, and significantly enhancing the intelligent interaction experience.
On the other hand, Geely's brother brand ZEEKR's intelligent cockpit team also announced the completion of deep integration of its self-developed Kr AI large model with DeepSeek-R1. ZEEKR Vice President Jiang Jun stated that the intelligent cockpit team will continue to train its own Kr AI large model's reasoning ability based on the DeepSeek-R1 large model to improve the understanding and prediction accuracy of users' vague intentions and implicit needs.
Following Geely, Voyah has also actively collaborated with DeepSeek in the intelligent cockpit area and announced the completion of deep integration. With the help of DeepSeek, the response speed, accuracy, and scalability of Voyah's in-car AI will be enhanced, and its "Xiaoyao Cockpit" will achieve diverse functions such as AI multi-semantic command recognition, AI poetry creation, AI painting, AI couplets, AI chatting, and AI real-time information retrieval.
It is understood that Voyah Zhiyin's in-car cloud will continue to integrate and distill the DeepSeek-R1 model to optimize the intelligent experience of the in-car system, achieve rapid technological iteration, and reduce the development and usage costs of the intelligent cockpit. Voyah Zhiyin will become the first mass-produced model to integrate DeepSeek, with plans for users to experience the AI intelligent cockpit through OTA updates on February 14th.
Additionally, IM Motors, Dongfeng, and other brands have also quickly seized the opportunities brought by DeepSeek and taken swift action. The announcement of multiple car companies joining in highlights the power of DeepSeek in the industry.
During this year's Spring Festival, DeepSeek saw a surge in downloads of 375% within just 7 days, topping the App Store's free list. Its performance, comparable to OpenAI's o1 model, has made it a sensation, with domestic banks, brokerages, technology, and automotive industries scrambling to board the DeepSeek fast train.
Currently, the competition in automotive cockpits has generally begun to focus on the level of intelligence, and DeepSeek, as the hottest topic in the large model field at the beginning of the Year of the Snake, entering the automotive circle is a natural progression. KaiYuan Securities pointed out that the R1 model is expected to bring a better cockpit interaction experience, and the cockpit Agent will achieve unprecedented functional enhancements In addition, many optimization methods of the DeepSeek-R1 model are expected to be referenced by the intelligent driving industry. KaiYuan Securities mentioned that current autonomous driving players are pushing large language models and even visual language action models to enhance the cognitive abilities of intelligent driving algorithms. DeepSeek-R1 is expected to serve as an excellent base model, distilling its performance to vehicle-side models and enhancing their capabilities.
It is worth mentioning that the reason DeepSeek-R1 can be implemented in vehicles about two weeks after its release is largely due to its powerful multimodal capabilities and the low-cost training methods and expenses.
Industry insiders believe that excellent user experience, faster iteration speed, and lower costs are also key indicators of competition among automotive companies in the second half of intelligentization, which will constitute a decisive advantage in the product strength of intelligent cockpits. Through relatively low performance overhead and training costs, DeepSeek has provided opportunities for many OEMs that cannot fully self-develop large models for in-vehicle use, promoting the democratization of AI technology.
It can be said that DeepSeek is expected to become one of the core selling points of smart vehicles, just like Qualcomm 8295 and NVIDIA Orin.
In fact, since the explosive popularity of ChatGPT launched by OpenAI in 2022, the concept of AI large models in vehicles has also become a hot topic. The Chinese automotive industry is quite determined in the direction of software-defined vehicles. Previously, with the involvement of large models like Wenxin Yiyan and Tongyi Qianwen, automotive companies such as Hongqi, Great Wall, Geely, and IM Motors have begun to promote the implementation of AI large model-enabled in-vehicle systems.
The explosive popularity of DeepSeek will undoubtedly accelerate the process of AI large models being integrated into vehicles, and analysts have referred to it as the "iPhone moment for automotive AI." This integration may quietly rewrite the way humans interact with steel machinery.
"In the next decade, AI will drive changes in automobiles that far exceed electrification," said He Xiaopeng, chairman of XPeng Motors, in a letter to employees.
Currently, the automotive industry is at a critical juncture in the second half of intelligentization. Players who can actively embrace this transformation and combine the technological advantages derived from DeepSeek with their own business are likely to stand out in the smart vehicle arena and occupy the high ground of intelligentization.
This fierce competition in intelligentization has also become a reflection of the intense rivalry in the automotive market.
In January of this year, there was a major reshuffle in the sales of new force car companies, with Hongmeng Zhixing and XPeng taking advantage of the situation, entering the ivory tower with sales exceeding 30,000 units; last year's strong performer, Li Auto, has temporarily retreated; while Xiaomi Auto ranked fifth with over 20,000 units sold; Nio and Leapmotor combined delivered less than 14,000 units, ranking lower among the top six new forces.
The market reshuffle is still ongoing, and the competitiveness among car companies is accelerating in differentiation. Yesterday's king may become tomorrow's has-been. Nio, XPeng, Tesla, and Li Auto have quickly launched various promotional policies, adding fuel to the competitive fire. At this time, the phenomenon-level DeepSeek has clearly been seen as a lifeline by many mid-tier and lower-tier car companies In the midst of an unclear and ambiguous landscape, all participants must grit their teeth and persevere.
Risk Warning and Disclaimer
The market carries risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investing based on this is at one's own risk