Xiaomi is rolling out a major software update for its electric vehicles, aimed at dramatically improving its Advanced Driver Assistance System (ADAS). The update is already live for the Xiaomi SU7 Pro, Max, and Ultra, and it will come preinstalled on the upcoming Xiaomi YU7 model. At the heart of this upgrade is a dramatic leap in the amount of real-world data the system has learned from. Xiaomi’s training material? Not just simulations or scripted scenarios—but 10 million clips of actual driving.
What Are Xiaomi’s “Clips”?
Unlike typical dashcam videos or social media snippets, these “Clips” are short 30–60 second bursts of real-world footage collected by the vehicle’s cameras, LiDAR, and mmWave radar. Each one captures a unique moment: navigating a congested street, responding to sudden braking, or squeezing past obstacles.
The system uses an end-to-end learning approach, meaning it doesn’t rely on fixed, rule-based instructions. Instead, the AI is trained by watching and mimicking human driving behavior. Think of it as a student absorbing millions of mini driving lessons, each taught by the road itself.
Xiaomi started with 3 million Clips. Now, that number has grown to 10 million, enabling the system to make smarter, more natural decisions.
Xiaomi ADAS update 2025: What’s Actually Improved?
According to Xiaomi, the difference is already noticeable:
- Smoother acceleration and braking — up 57% in ride comfort
- Obstacle navigation — improved by 67%
- Intersection handling — up 23% in success rate
This should mean fewer jerky stops, better maneuvering around double-parked cars, and more confidence at messy intersections. The aim is clear: to make assisted driving feel more human, less robotic.
Massive Investment, Bigger Ambitions
This update is part of a much larger push by Xiaomi into the smart vehicle space. The company has poured ¥5.79 billion (roughly €691 million) into its EV division. More than 1,800 engineers are on the ADAS team alone, and they’re backed by over 400 test cars collecting live data from public roads.
Powering the AI is MiMo, Xiaomi’s proprietary large language model, built inside its in-house AI Lab. This is what makes it possible to process the massive Clip dataset and turn it into real-time decision-making for the road.
A Step Toward Autonomous Confidence
For Xiaomi, this isn’t just a software patch—it’s a proof point. A signal that its smart driving tech is maturing fast, backed by serious data, serious funding, and serious results.
If you’re behind the wheel of an SU7 or soon-to-own a YU7, chances are your car is learning to drive like a pro—10 million lessons at a time.
