Beijing traffic does not care about your feelings. That's where the lane line is simply decorative. The driver there moves with a level of aggression that makes the taxi driver in New York City blush. When the car hesitates for a little while, the other 3 cars fill that space. This is the ultimate testing ground for electric cars that claim they can drive themselves. I went to Beijing to test the new XPeng VLA2.0 and witnessed it go on these chaotic streets. The results suggest that the era when Tesla is the only serious player in the autonomous driving game is officially over.
XPeng began sending VLA3 software through the over-the-air update in 2026/2.0. This directly applied to the "Ultra" versions of the company's Xpeng P7, XPeng G7, and XPeng X9. VLA stands for Vision-Language-Action. The 2.0 version cam up with some big chnages on how Evs understand the world. The old version had some draw backs such as, one part looked at the road, another planned the route, and the third moved the wheels. This often caused delays and jerky movements. VLA2.0 uses a single "brain" that looks at the camera image and instantly determines the action. It removes middlemen, making the driving feel much more natural.
To make this possible, the company has built a very powerful dedicated computer chip - Turing AI. It can handle above 2,250 in computing power. Then XPeng trained the system with 100 million video clips of the most difficult driving moments imaginable. While test driving for around 40 minutes in Beijing, the system made all its decsion without any human input. In a city where people shut you off every 30 seconds, the car was calm and moved through busy intersections dealing with aggressive drivers without getting stuck as if it were a gentle walk in the park. There was no need to grab the wheel or brake to save the car from mistakes. This level of reliability is rare in the modern Ev world. One moment during the drive showed exactly how smart the system became.
The car had to megre on a very fast moving traffic where the gap was very narrow. Most driver assistance systems in other electric vehicles will be confused here. They usually wait for huge openings that never come, or give up and ask humans to take over. XPeng P7Ultra did not do either. It saw a small gap, calculated the speed of the other car and moved firmly. It claimed itself in the same way as the local driver. It was a smooth move that showed that the car had real confidence. In another situation, the car came out of the country road to the main 3 lanes through the road with traffic coming from 5 different directions. At least 30 scooters were trying to push back into the country road, there were pedestrians navigating this madness, cars coming from the left and from the other side of the road.
Read More: 2026 DS N8 AWD Review: French Luxury Meets Electric Performance
Honestly, it was a challenging situation for experienced drivers with well-flowing traffic beyond the legal speed limit. Still, the P7 quietly nudged forward and positioned itself for the innermost lane, punching the moment it saw a gap large enough to participate in traffic. While driving out of Beijing, we approached the blind bend with yet another scooter trying to avoid bumps and potholes. The P7 slowed down early, retreated and actually approached the right edge of the road. This is what experienced drivers do to see the details of the road, turn left ahead - give the scooter rider enough space.When the car is sure there is no other traffic, it is bolted forward, giving the rider a wide berth and completed the overtake in seconds. The experience was strange.
It was like seeing a driver doing it for years. It was really impressive to see, without hesitation, fully aware of everything that is happening around, enough room left for everyone on the road, and not changing minds and stopping. Another interesting situation was looking at the location of the car itself in the middle of the road. P7 continued it on narrow roads without marks, the edges are uneven. Instead of strictly sticking to the right edge, steer away from all the bumps and come back only when you see the approaching traffic. This is also what experienced drivers do to stay away from potential hazards and leave room for reaction in the event of an emergency.
All this human-like confidence comes from a "generative world model." This is a virtual reality world where xpeng can test its software without hitting a real road. The company is stepping up testing from 30,000 scenarios per year to 500,000 scenarios. Every day, this system "drives" the equivalent of 3000 million kilometers in this digital world. By doing this cars will learn from the mistakes of the simulation and will make sure that no such mistake shall happen in the real world. On the other hand if we talk about Tesla, they heavily rely on data from real cars on real roads, but XPeng uses these simulattors to teach its cars how to work
He Xiaopeng, CEO of XPeng, recently visited Silicon Valley to test Tesla's FSD v14.2. He spent five hours driving around San Francisco and actually admired the system. He also set very tough goals for his own team. He wants the VLA system to match Tesla's best performance by 2026/8/30. Owners in North America use the latest software, while Tesla drivers in China still use older versions. The Chinese government has been slow to approve Elon Musk's company's latest autonomous driving feature. This will allow local brands to make a big start. Companies like BYD have already installed the "God's Eye" system on EVs, which cost about 26,000 euros.
Read More: 2026 BMW iX Is the Luxury EV That Quietly Improves Everything
Meanwhile, the tech giant Huawei has spent more than 90 billion euros on its own driving software. Even Xiaomi is jumping into the ring with its SU7 lineup. The market is very crowded and the technology is moving faster than most people can keep up. The business aspects of these autonomous systems are also changing. Tesla charges a subscription fee of €85 per month for full autonomous driving capabilities in the United States. For now, XPeng has a VLA2.0 system included in the price of the car. This makes it very difficult for Tesla to justify its monthly bill if its competitors are offering similar things for free. Volkswagen seems to agree that XPeng has become something special.
The German automaker has signed a contract to use VLA2.0 on a new electric SUV built specifically for the Chinese market - a huge vote of confidence from one of the oldest names in the automotive business. But I believe VLA2.0 feels different because it predicts what will happen next. Unlike many other systems already on the market, it does not simply react to a car stopping in front of it. It sees the flow of traffic and places itself early for turns and merges. This eliminates the "not-so-sure" thing that has plagued many other driver-assist systems. Many cars feel like they're guessing what to do and make passengers nervous.
This system feels like there is a plan. It quickly and safely corrects mistakes even when small errors occur, such as getting too close to the car or hesitating on the turn. The automotive scene is really at a turning point. People used to buy Tesla because it was the only "smart" car available. Currently, the XPeng P7, XPeng G7, and XPeng X9 prove that other companies can build just as sharp brains. The new XPeng GX promises even more with a whisper of level 4 autonomy. Whether XPeng is officially "better" than Tesla is an issue that requires more side-by-side testing. But for those driving the Beijing traffic madness, the VLA2.0 system feels like a very capable partner. And if it can do it safely in Beijing, it should not have problems dealing with traffic anywhere else in the world. Honestly, the gap is closed and the race for the future of EVs is now dead heat.
Read More: New Audi RS5 Turns Heads With Its Shockingly Powerful Plug-In Hybrid V6
Comments
Post a Comment