Take a Chinese car to the other side of the world and the trouble is often not the engine, but the mouth. You want the car's system to find a parking lot or adjust the air conditioning, but it only understands Mandarin; the street names the navigation reads out are spelled in ways you never learned. The person has gone abroad while the car stays behind, an awkwardness familiar to many who drive overseas.
This is now being fixed. According to China Automotive News, at iFlytek's Global 1024 Developer Festival in October 2026, the Spark Smart Cockpit 3.0 appeared, built precisely to let the car's system understand many languages and finish the task. It went into mass production as soon as it was unveiled, rather than remaining a concept piece on a stand.
Bring the lens back home and the same energy is not used only in cars. According to Xinhua News Agency, on November 2 the Light of the Internet Expo of the World Internet Conference opened in Wuzhen, with more than a thousand new technologies and products on show; smart cockpits, AI glasses and digital workers are all walking the same road into real scenarios.
An Unfamiliar Accent Behind the Wheel
For a car to be sold overseas, its cockpit first has to learn how the locals speak. Voice commands, navigation announcements and menu text all have to change with the market, and accents and habits of expression matter too. In the past this adaptation was done by people one model at a time: slow, and hard to cover thoroughly.
According to China Automotive News and Zhong'an Online, the cockpit is driven at its base by the SparkAuto automotive large model through cloud plus on-device collaboration, and its first partner brands include Chery, Great Wall's Wey brand and Dongfeng Nissan. Note that this is a first list of partners, not coverage of every carmaker, and should not be read as more than it is.
Several Wishes in One Sentence
What it wants to solve is not merely turning one sentence into another. According to China Automotive News, as a Master Agent multimodal hub, it can understand several requests made at once, weigh constraints such as time, budget and parking, reorder the tasks automatically, recommend a suitable place, and, once you confirm, call up navigation to carry it out.
This is very different from older car systems that moved one step per spoken command. Say you want to eat, charge the car on the way, and still arrive before six, and the system has to put the several tasks in order first, then pick a place that is on the route. It manages a chain of actions rather than one instruction, which makes the cockpit feel more like a co-pilot who can make arrangements.
There is another easily overlooked change in the cockpit: the system no longer waits for you to break your needs into separate commands, but can take in a whole string of vague talk. Understanding, ordering, confirming, executing: with those four steps linked, it feels like a person helping, not a machine waiting for a password.
Heard Both Inside and Outside the Car
There is an entry point outside the car as well. According to China Automotive News and other reports, the system supports long-range voice wake-up and voiceprint recognition, so when your hands are full of bags or you are circling a parking lot looking for the car, you can call or control it by voice from dozens of meters away. The voiceprint confirms who is calling, so a stranger nearby cannot simply summon the car.
None of these scenes is striking, yet they are exactly where drivers most often get stuck. No free hand, a screen you cannot see, forgetting where you parked: in the past you just dealt with it; now a single spoken call is caught, and one layer of trouble disappears. Whether a technology is good often comes down to whether it bothers with these few steps.
Cloud and On-Device Working Together
Why insist on cloud plus on-device? According to China Securities Journal, the Spark X2.6 follows a device-cloud integrated route: simple matters and those involving privacy stay local, complex ones go to the cloud, and where cooperation is needed the two work together. Only if device-cloud routing judges accurately will tasks not be sent to the wrong place.
One layer further down is computing power. According to China Securities Journal, Spark X2.6 was trained entirely on a domestically built computing platform, and the Feixing platform was built by iFlytek together with Huawei on the Ascend ecosystem. With an independent base, the cockpit's multilingual capability can keep improving, rather than being a one-off adaptation.
The device-cloud route was not a sudden idea. According to China Securities Journal, the company disclosed earlier that device-cloud routing accuracy had risen from just over 80 percent to 96.1 percent, and that the Spark X2.5-4B on-device model was downloaded 2.56 million times in its first month after being open-sourced. These figures come from the company itself; read them for the direction: models are getting smaller, able to fit into cars, glasses and computers.
In a car, this division of labor is easy to understand. In a tunnel with no signal or an underground garage with weak coverage, the on-device side has to hold wake-up and common commands steadily; when it comes to planning a cross-city route or checking a pile of real-time information, the cloud takes over. Putting computing power in the right place is more useful than simply piling on parameters.
The same base is also laid elsewhere. According to the Workers' Daily app and Enorth, iFlytek's medical imaging cloud covers more than 260 million people nationwide, and a single CT scan can flag multiple conditions. From classrooms and clinics to cockpits, the domestic computing power and model capabilities underneath are similar; only the shapes they grow into differ.
Going Overseas with Chinese Cars
On going overseas, public materials give one set of figures. According to China Automotive News and Zhong'an Online, iFlytek's automotive multilingual technology already supports 32 overseas languages and covers 60 countries and regions; among China's top ten car exporters, eight work with iFlytek. These are figures disclosed by the company, and what they show is breadth of coverage.
Multilingual capability was not a lesson crammed overnight either. According to the company, iFlytek has provided simultaneous interpretation and transcription for large international conferences for years, and the corpus and acoustic models accumulated there have in turn fed this cockpit.
Seen another way, this is a trade line that gets little attention: wherever Chinese cars are sold, the Chinese AI inside them follows. It does not go overseas on its own, but hides behind the steering wheel and, car by car, quietly enters other people's daily commutes.
Nor should the story be told as complete. The first partner brands and the coverage figures the company discloses are still moving forward; whether a given language is translated accurately, whether a street name is pronounced right, has to be worn smooth by real road conditions. Going overseas is a long run, not something one launch event finishes.
The Same Foundation Underneath
Beyond language, data compliance, local regulations and operations systems all have to be built country by country. Multilingual support is only the ticket; whether people in different countries find it smooth and trustworthy is the real homework of going overseas. None of this work appears on the launch-event screen, yet it decides whether a car can keep being sold over the long term.
The bustle of the launch event will eventually fade. What remains is an ordinary morning abroad, when someone gets into the car, says a sentence casually in the local language, and the car understands and gets it done. No applause, no demo, just a smooth stretch of road.
For ordinary people, the standard for judging whether this kind of technology has landed stays plain: has it taken over those unavoidable little troubles? Not understanding, not finding, not having a free hand all count. One fewer moment of fumbling is worth more than one more line on a spec sheet.
Driving overseas and hearing the car speak the local language sounds like a light detail. Yet it is exactly these details that decide whether a technology stays on the stand, or truly follows people out of the country and into their days.
Disclaimer: This article is based on publicly available information and is for reference only. It does not constitute investment advice.