As new energy, autonomous driving, artificial intelligence, and chip technology rapidly evolve, the automotive industry is undergoing an unprecedented transformation. To stay abreast of the latest trends and market dynamics in the automotive sector, LingTai LT has launched the "Auto Circle" column. From a professional standpoint, this column delves into the latest developments in the global automotive industry, new product releases from leading automakers, technological innovations, and market performance. Through in-depth analysis, it uncovers the underlying business logic and market patterns, as well as how they are reshaping the industry and influencing human mobility. This article, the 34th installment in the column, focuses on the commercial strategy behind Tesla's decision to integrate the Doubao large model.

Author: Ma Duoduo

Editor: Hu Zhanjia

Operations: Chen Jiahui

Produced by: LingTai LT (ID: LingTai_LT)

Header Image: Publicly available online image

There was no launch event, no appearance by Musk, and not even a formal press release.

On August 19, Volcano Engine officially confirmed that Tesla China's in-vehicle system had officially launched the Doubao large model, with features being rolled out in batches. Twenty days prior, on July 31, Tesla China began rolling out the OTA update version 2026.14.13 in batches across all Model 3, Model Y, Model S, and Model X variants. The Doubao large model was quietly installed as a standalone app in the in-vehicle system of this American company, renowned for its full-stack AI self-research capabilities.

This marks the first time in over a decade since Tesla entered the Chinese market that it has integrated a third-party large model into its in-vehicle system. For a company that emphasizes self-research as its core philosophy, this move carries significantly more weight than a routine software update.

A Year-Long Prelude to 'Integration'

A chronological review of publicly available information reveals that this collaboration has been approached with considerable caution and patience.

According to comprehensive media reports from Tianyancha, in August 2025, Tesla reached a strategic cooperation agreement with Volcano Engine. At that time, Tesla's official "Terms of Use for In-Vehicle Voice Assistant" already outlined a dual-model solution. In April 2026, Tesla's in-vehicle voice large model service completed registration in Shanghai. By the end of July of the same year, the OTA update was officially rolled out in batches, with newly delivered vehicles given priority. From signing the agreement to integration, the process took a full year.

The final solution implemented is a dual-model synergy: Both the Doubao large model and DeepSeek are accessed through Volcano Engine in the in-vehicle system. According to the officially disclosed adaptation plan, Doubao handles vehicle control-related voice commands, such as navigation settings, media playback, and air conditioning adjustments, as well as owner's manual queries. DeepSeek, on the other hand, is responsible for open-ended interactions like casual chatting, weather, and information queries. One model handles tasks, while the other handles chatting—a deliberate division of labor. The accuracy required for vehicle control commands and the openness needed for casual chatting are difficult to reconcile within a single model.

However, the initial experience for the first batch of vehicle owners falls short of the planned solution. Media tests by Pacific Technology and others show that in the current rolled-out version, Doubao has not yet been granted full vehicle control permissions. Some owners report that it is "equivalent to adding a chatty app to the in-vehicle system." Additionally, there are two significant barriers: This feature requires a monthly subscription of 9.99 yuan for advanced in-car entertainment services and is only compatible with newer models equipped with AMD chips. Older models with Intel chips cannot be upgraded due to insufficient computing power. Tesla's localization and delegation of authority are proceeding step by step.

Why Did the Self-Research Philosophy Make an Exception in China?

To understand this "exception," one must first recognize the three practical constraints Tesla faces in China.

The first constraint is compliance, an unavoidable barrier. In-vehicle large models in China must undergo registration, and data must remain within the country. However, Grok's servers are located in the United States, and direct integration into Chinese in-vehicle systems would involve cross-border transmission of sensitive data such as voice and location—a threshold that cannot be circumvented.

The second constraint is language itself. Grok's training is primarily in English, whereas Chinese spoken language involves elliptical expressions, ambiguous commands, and dialect recognition, all of which rely on long-term accumulated data from local scenarios rather than sheer parameter scale.

The third constraint is the competitive pace. While NIO's NOMI and Li Auto's "Li Auto Classmate" have elevated cabin interactions to emotional, multi-round dialogue, and whole-vehicle orchestration levels, Tesla's voice system remains stuck at the command recognition stage. The amplification of this shortcoming through comparative consumer decision-making has occurred much faster than expected.

By choosing Doubao, Tesla gains more than just a model.

According to comprehensive media reports from Tianyancha, Volcano Engine has previously collaborated with mainstream domestic automakers, with over 7 million smart vehicles equipped with the Doubao large model, covering more than 50 brands and 145 models, and completing over 30 million daily cabin interactions. For Tesla, which is meticulous about delivery quality, this automotive-grade engineering capability, validated across millions of vehicles, is just as important as the model itself.

Third-party data shows that the Doubao app has over 380 million monthly active users—integrating a model from a nationally popular app into vehicles effectively leverages existing user habits in one fell swoop.

The Real Battle Lies Beyond the Cabin

While the cabin voice shortcoming can be quickly addressed through collaboration, Tesla's true competitive edge in China does not lie here.

The process of FSD's entry into China has been a tortuous five-year saga: In February 2025, it briefly rolled out assisted driving functions for urban roads but suspended them a month later due to regulatory approvals. In November 2025, Musk optimistically predicted approval by early 2026, but by the Q1 2026 earnings call, the CFO adjusted the timeline to Q3. On May 21, the supervised version of FSD was officially announced for the Chinese market, but as of late May, it had only been rolled out in a gray release to about 5,000 vehicles equipped with HW4.0 hardware. More critically, its positioning is strictly defined as Level 2 assisted driving, requiring the driver to remain fully engaged and legally responsible. Tesla China's official website has revised related descriptions from "fully autonomous driving" to "assisted driving."

Pricing is another hurdle for FSD.

Its one-time purchase price in China has long been set at 64,000 yuan, while Huawei's ADS 4.0 costs users 12,000 yuan after subsidies, and Xpeng and Li Auto's solutions are priced between 30,000 and 36,000 yuan. As domestic intelligent driving systems accelerate their popularization, and urban NOA (Navigate on Autopilot) penetration rates continue to rise, the value of FSD's "futures ticket" is being diluted by time. According to multiple media reports, the full-featured version of FSD may begin a broader rollout in late August, but whether it will pass regulatory approvals and how many models it will cover remain subject to official information.

While the bottom card of intelligent driving is repeatedly delayed, market pressures are concrete.

In Q1 2026, Tesla China's retail sales reached 112,800 units, down 16.2% year-on-year. For the full year of 2025, Tesla's global deliveries stood at 1.636 million units, surpassed by BYD's 2.26 million units, losing its title as the global pure electric sales champion. Although Q2 showed signs of recovery—according to the China Passenger Car Association, Tesla China sold a combined 89,000 Model 3 and Model Y units in June, up 24.4% year-on-year, and Shanghai Gigafactory delivered nearly 468,000 units in H1 2026, up 28.4% year-on-year—the intensity of competition in the Chinese market is nothing like it was in 2019 when the Shanghai factory first started production.

Collaborating with Doubao to Adapt to Local Customs

On July 30, Tesla celebrated the rollout of its 10 millionth pure electric vehicle globally at the Fremont factory, with nearly half of them produced in Shanghai. The next day, China's in-vehicle system integrated a Chinese large model. These two adjacent days mark the dual nature of Tesla's relationship with the Chinese market: deep reliance on China for manufacturing and increasing reliance on China for intelligence.

For Tesla, integrating Doubao does not mean abandoning its self-research route—overseas in-vehicle systems remain Grok's domain, and FSD will not cede control to anyone. What has truly changed is how multinational companies operate in the Chinese market: For user-centric aspects like cabin experience, content ecosystems, and application services, local solutions have become the lowest-cost and fastest-to-implement choice. Meanwhile, self-research barriers for core technologies continue to be firmly held. According to Zhejiang News, tests for Alibaba's QianWen large model in Tesla's in-vehicle system are also underway. Having multiple Chinese large models running in a single vehicle may soon become the norm.

From the Shanghai factory addressing the question of "can it be built" to Doubao's integration answering whether Chinese consumers "like it," Tesla's localization has progressed through two stages.

As for the third stage—whether FSD can truly land and at what price—that will determine Tesla's next five years in China. This time, how long Musk can wait is not up to the market or regulatory approvals but ultimately up to Tesla's Chinese competitors, as reflected in the monthly sales rankings.

For inquiries about the article, please contact

Follow LingTai's video account to stay closer to business!