Chinese NEVs Enhance Intelligence with DeepSeek Technology

Are you all using DeepSeek? Since its breakout, everyone want to have it and use it. Chinese automakers are rushing to integrate it on their lineups. Until now, there are over 10 Chinese automakers announcing that they have already or are going to integrate DeepSeek-R1 model into their lineups, covering over 20 brands.

Here is a list for your reference.

Geely Auto

Models/Systems Integrated with DeepSeek: Geely StarRay Large Model, and Zeekr Intelligent AI assistant Eva

Equipped Models: Models to be equipped with StarRay Large Model, and models from Zeekr

Functions Improved: To understand the commands precisely, access to 2,000 in-car interfaces, and analyze user needs proactively.

Dongfeng Motor

Models/Systems Integrated with DeepSeek: Voyah Xiaoyao Cockpit, and Cockpit applied by models under M-Hero

Equipped Models: Voyah Zhiyin (Courage), Voyah Dreamer, M-Hero 917, Dongfeng Yipai (eπ) , Dongfeng Fengshen (Aeolus), Dongfeng Nammi  

Functions Improved: To recognize multilingual command, retrieval in real-time, interact proactively with human.  

Changan Auto

Models/Systems Integrated with DeepSeek: Dubhe Plan 2.0 and the Qiyuan Deepal  Vehicle computer system

Equipped Models: Qiyuan E07 and all models from Deepal  

Functions Improved: To help Dubhe Plan 2.0 achieve comprehensive AI intelligent driving capabilities, and to improve the speed and logic of voice interaction.  

SAIC Motor

Models/Systems Integrated with DeepSeek: SGMW Baojun Lingyu Intelligent Cockpit, and IM Motor Intelligent Cockpit

Equipped Models: Bojun Xiangjing, Baojun Yunhai, Baojun Yueye Plus, and IM L7、LS6  

Functions Improved: To response simple queries quickly, and deeply process complex needs.  

Chery

Models/Systems Integrated with DeepSeek: Xiongshi Intelligent Cockpit

Equipped Models: Chery, Exeed, Jetour, and iCar  

Functions Improved: To improve the accuracy of the response to the voice commands, and make the human-vehicle interaction smarter.    

BYD

Models/Systems Integrated with DeepSeek: Xuanji Architecture

Equipped Models: Han and Tang models  

Functions Improved: To enhance vehicle and cloud AI capabilities, accelerate smart driving training efficiency, improve the logic reasoning abilities and optimize the decision-making abilities.  

FAW-VW

Models/Systems Integrated with DeepSeek: New Media AI Content Platform

Equipped Models: None  

Functions Improved: To assist sales in marketing content generation and precise delivery.

Smart China

Models/Systems Integrated with DeepSeek: Self-developed AI Agent

Equipped Models: Smart series  

Functions Improved: To optimize multilingual interaction experience  

Leapmotor

Models/Systems Integrated with DeepSeek: New Cockpit System

Equipped Models: Unknown  

Functions Improved: To improve the ability to solve complicated problems.  

GAC Group

Models/Systems Integrated with DeepSeek: ADiGO SENSE System

Equipped Models: ADiGO SENSE models（e.g. Aion series）  

Functions Improved: To improve the ability to understand commands and response speed.

BAIC Motor

Models/Systems Integrated with DeepSeek: Intelligent Cockpit

Equipped Models: Arcfox S5 model and more  

Functions Improved: To improve the experience of human-vehicle interaction.  

From the above information, we can see three applications in auto manufacture:

- Most of the other automakers are combining it with their in-house developed smart cockpits.

- BYD and Changan appear to integrate DeepSeek in their advanced intelligent driving technology and smart in-house developed smart cockpits.

- FAW-VW used it as a tool to improve new media content, thus to boost sales.

Basically, as we know, there are two ways making cars smarter:

-  AI into the smart cockpits: to have a smoother and more humanized human-vehicle interaction experience.

-  AI to enhance advanced intelligent driving: improved the intelligent driving applications, especially on L3/L4 autonomous driving technologies.

According to the anticipation of the third-party institution, it takes traditional automakers at least 3 to 6 months to train their autonomous driving models, while by utilizing DeepSeek's MoE architecture, the time will be shorten to two weeks. It is believed that we will see more cases where DeepSeek can help improve the advanced intelligent driving capabilities of cars.