Tesla FSD (Full Self Driving version 13.2.6) has been localized to China using only internet driving videos of driving in China.
Previously it was believed that Tesla would need to be able to send a lot of driving videos collected in China to retrain the base model. This was not needed. There was no regulatory change needed for sending data out of China.
This means localization is easy for tesla and it’s AI. Everywhere in the world with no data transmission issues. Like Europe. Unsupervised robotaxi will be global. This will further improve this year with Grok3 , Grok Video and Grok 4.
Those systems will be even better at learning from video and understanding other languages.
There are now driving videos from all over China where there are happy and impressed drivers using the driving system on city streets and highways.
This improved capability will increase sales of Full self driving software in China and it will increase the sale of tesla cars.
A separate report is that the new refreshed Tesla model Y production has reached 2000 cars per day (14,000 cars per week). This is a significant ramp and reports are new model Y production at Fremont and Austin are going well.
Tesla FSD in Chongqing, China (dubbed into english) $TSLA pic.twitter.com/oKS3OJTMfe
— Whole Mars Catalog (@WholeMarsBlog) February 26, 2025
Another Tesla FSD video from the busy streets of Putian, China! 🇨🇳
The rollout is happening, and it’s impressive to see! 😎$TSLA
— Herbert Ong (@herbertong) February 25, 2025
$TSLA 🇨🇳 You may wonder HOW FSD PERFORMED IN CHINA AT NIGHT? Here's a video which can give you a hint (with ENG caption)
Overall Impression (Good with some limitations):
– The FSD performance exceeded the user's expectations, even in challenging conditions.
– Object detection… pic.twitter.com/cyufezVdRK— Tesla Α & Ω (@tcm1907) February 25, 2025
Yo, China roads are WILD
FSD in China performs flawlessly in this video.
One can't help but smile watching this. pic.twitter.com/qS0zAkUzQq
— Tesla Hype (@TeslaHype) February 25, 2025
Chinese blogger’s view of FSD in China so far:
“Tesla FSD unexpectedly went live today, and it behaves better than I anticipated.
Its core capabilities are impressive, with no strict ODD (Operational Design Domain) limitations. Some owners have successfully used it in underground parking lots, set random destinations without issues, and executed U-turns flawlessly—features we’ve been testing and gradually rolling out.
What’s remarkable is that this has been achieved without a large-scale test fleet, without an extensive local development team, and possibly even without concentrated local data training.
The failures some media outlets predicted—being unusable or causing frequent accidents—simply haven’t happened.
The driving feels remarkably human-like, and its overall planning capabilities are on par with the best in the field.
This suggests that, with vast amounts of training data, even data sourced from the U.S. and relying solely on a vision-based approach, Tesla’s system has developed exceptional transferability.
That’s truly astonishing.
In contrast, most of our domestic end-to-end teams tend to overfit results to specific cities. A nationwide rollout of end-to-end systems is still some distance away—let alone achieving cross-regional adaptation.
Of course, occasional issues like running red lights, going straight from left-turn lanes, or drifting into the wrong lane still occur.
I’ve mentioned before that some abilities are foundational, while others depend on learning specific rules. Even without fully mastering those specific rules yet, Tesla FSD’s fundamental capabilities have already proven themselves.
It can be deployed across different regions with virtually no additional software development costs.
Whether it can truly master those specific rules now depends on whether Elon Musk is willing to invest—and how he chooses to do so.”
Chinese blogger’s view of FSD in China so far:
“Tesla FSD unexpectedly went live today, and it behaves better than I anticipated.
Its core capabilities are impressive, with no strict ODD (Operational Design Domain) limitations. Some owners have successfully used it in… pic.twitter.com/2NO7cFJcHG
— Ray (@ray4tesla) February 25, 2025

Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
Oh my God, this MUST have been SO interesting. If I may. I “define” an interesting event that happened many years ago. I hosted a gentleman from the nation of Eritrea who asked a very uhm, interesting question. This guy was a senior member of that government at the time. The question asked, was interesting. He asked: “When four cars, come to a four way stop, no lights, no stop signs, who knows who should go first? I said under uncontrolled conditions, the first person to the intersection goes first, the second goes second, and so it goes. Not complicated.
I said, we all want to get before the end of the day, in some place, in one piece. We obviously all want to do THAT. IMO, it’s not a matter of what’s right or correct at any moment anywhere and so called everywhere. It’s what works, at anyone place at anyone moment in time. Deal with it gang. In one country, you need to be in a tank. In another, you just need to know and follow the rules. And the rules will work for you. Ain’t that the point?