First Perfect-Score Contender Emerges! Aion i60 & WeRide Break New Record in "Smart Exam Peak Hour"
Combine 50 kilometers, 20 challenging scenarios, and rush hour traffic, and the difficulty is roughly equivalent to turning a typical urban commute route into the Nürburgring.
The model that was showcased last week is the i60 from Aion, priced under 150,000 RMB.
The car runs L4-derived algorithms - WeRide WRD 3.0, with the chip coming from the Qualcomm Snapdragon 8650, delivering 200 TOPS of compute power.
In fact, its paper strength doesn't stand out.
Here's the result: It broke the best record of Huawei QiankunADS 5, which had held for two months on this route, and is also the first vehicle to fully pass all 20 challenging scenarios.
But on the other hand, it's also important to note that "passing without damage" doesn't mean there are no weaknesses. The i60 still engaged 3 times throughout the entire process.
Here's the Aion i60's test score : 20/20, 3 interventions.
Putting these two numbers together reveals the most interesting aspect of this test – it could end up being this year’s best score, while also clearly crossing the line in a few specific judgments.
This article starts by looking at these three interventions, and then examines how this system achieved this result.
Three interventions categorized by nature: two "efficiency mistakes" and one "didn't read the road". We classified the three interventions into two categories: one where they knew the rules but tried to go ahead for efficiency, and another where they missed understanding the long-tail scenarios.
It first happened on a congested road.
There are about 450 meters left to turn right, and the right side is the bus lane, which is currently in a no-entry period.
So everyone is orderly lining up, and no one is taking any unnecessary risks.
Just after I60 passed a crashed vehicle, it abruptly turned on its turn signal and merged into the bus lane directly.
In reality, sometimes right-turn-only lanes and bus lanes overlap, and many drivers will cut in directly at the last few meters; but seeing the temptation of hundreds of meters of empty road ahead, we need to take it.
It was no different; it wouldn’t give up even after failing once, and after taking just a few steps, it was already thinking about entering the bus lane.
That's not unjustified at all for a traffic violation.
Oddly enough, at the same time and location, iCAR V27, which was equipped with Horizoond's HSD 2.1, also made the same mistake.
This highlights an interesting point: the i60 knows exactly when it can use bus lanes, but its relentless pursuit of efficiency can occasionally lead to it making mistakes.
The next two interventions were less about not understanding the tail-end scenarios and more about whether the system can read sky signs and whether it dares to enter a tidal lane.
This is not difficult for many ADAS systems.
To improve traffic efficiency, the i60 will also enter the variable lanes; however, as things develop, a huge bug is waiting for it: at the end of the lane, the double yellow lines disappear, and there are no guide lines on the ground.
Continuing straight seems like no problem for ADAS (Advanced Driver Assistance Systems), but going forward would be backwards.
The "disappearing lane" trick left many cars, including the i60, in a bind, ultimately leading to a violation.
The last intervention occurred during our classic rail segment.
The priority was with the stop sign before the railroad tracks. The answer is the stop sign.
Although there are changing traffic lights at this intersection, whether one can proceed depends on whether a train is passing through, so the priority goes to the poles, and the traffic lights are basically just for show.
The green light and the stop sign both went up, and the car in front simply went through the intersection.
When it was i60's turn at the green light, there wasn't enough time, so it drove forward a bit and found a "stop line" to stop at the intersection.
But this stop line – it seems like a rail.
It determines the priority of the rules, but the spatial semantics are still one layer away.
Looking at them together, you can see that none of the three interventions were due to a fundamental capability failure.
Bumping into bus lanes is efficiency trumping rules, misinterpreting the end of a tidal lane as a tail-end understanding, and misreading the meaning of railway crossings.
These are 'boundary issues', not 'whether you can drive' issues.
These situations, without intervention, don't necessarily mean they were handled well. There are also a few scenarios where, while we didn't intervene, we feel that the handling of i60 could have been smoother.
First, it occurred during a straight-line drive.
When the car in front flashed its right turn signal, the i60 immediately initiated a lane change, handled very smartly.
But what really stumped it was a subsequent accident.
We thought the Zeekr 9X on the right lane was about to turn right to merge, but in reality, it was forced to stop in a very awkward position due to an accident.
i60 already shows an awareness of lane changes through 360, but when the traffic starts moving, it gets stuck in this simple lane change.
Embarrassment won't disappear, it'll just shift.
The i60 drew a "me fence."
To alleviate the embarrassment, the driver and navigator had been explaining what was happening; once traffic started moving again, the i60 was finally able to get unstuck.
This incident occurred at a crossroad.
When the signal was about a hundred meters away and turning yellow, the system immediately recognized the change in the lights.
But due to the high speed at that moment, combined with the stop line appearing much earlier than expected, we managed to brake and receive the only alert of the entire journey.
The issue is in the margin of decision-making.
Third, it's a right turn.
The black car in front charged ahead, cutting through the solid line and taking half of the bike lane.
The i60 follows aggressively.
Initially, they thought it was just hitting a solid line, only to realize they had been completely misdirected upon the driver's realization.
The commonality among these three is that it knows what it should do, but it is not decisive enough in its actions and the distribution of reserves is insufficient.
For everyday use, they won’t directly turn into a danger, but they will make people sitting in the car noticeably feel that “it could be better.”
Lastly, I want to emphasize one thing about the system's feel.
I admit it's a bit of a drag, that feeling of pushing back against it.
For example, if you want to reverse a lane change, the most intuitive way is to steer the wheel back, but the i60's wheel gives you a strong and persistent force, as if you're wrestling with a ten-year-old child.
Such a tactile response can easily lead to driver confusion and misoperations operations in many emergency situations.
Strongly recommended for optimization.
The more capable human-machine collaboration becomes, the more important the "exit mechanism" becomes.
A system that can drive a car well should also allow for a clean and smooth handover – and these are not contradictory, but complement each other.
But the standout of Wen Yuan WRD 3.0 can be summed up as 'fast, smooth, and agile'. After discussing the capabilities and areas for improvement of this system, let's now talk about the features and highlights of the i60 overall in operation.
Fast, Smooth, Agile
Let's go fast.
This time, I truly experienced what it means to drive like in a game of "Asphalt".
Take the late peak hour congestion on the Fourth Ring Road, for instance. Instead of merging by alternating between driving forward and looking for opportunities, the i60 directly implements a pit-style lane change – by the time the cars ahead are still looking for opportunities to merge, we've already pulled into the lane.
Right there is another car with adaptive driving assistance and its blue turn signal, what looks like a showdown between regular blue lights.
It's up for debate; you decide who wins and who loses.
The same move is used when shifting out of the main road, and it works every time.
Pay attention: It seemed like the right lane was moving faster at the moment, so the system decisively initiated a lane change. Only to realize halfway through that the original lane was much clearer, so it immediately changed back.
It's not a lack of confidence, but rather the choice made after taking a glance.
It's getting undead.
If the Quadra system was too congested earlier and limited its performance, then try looking at its efficient operation at a different time: the i60 immediately went "crazy" as it darted in and out, not staying idle for long, and even managed to overtake the other cars with blue lights on the ADS.
Lastly, looking at a right turn: facing the two-wheelers coming from all directions, the i60 keeps them in sight and in mind. After carefully controlling the distance and finding the right timing, it gradually starts to press forward.
Agile style and prioritizing efficiency, it's not just for show.
Use with caution for daily use.
The second point is alignment.
What's the deal? The first point is what I mentioned earlier: It's hard to remove it.
The ADAS system has an almost obsessive grip on this car, leading drivers to rarely intervene, which results in a nearly 100% activation rate, providing a smooth and seamless experience with no lags.
Second, the car's acceleration and deceleration control is extremely fine-tuned.
Let's take the last scenario as an example: on a tight corner on a ramp, very few drivers can handle the corner smoothly, and it's difficult to control the deceleration before entering the corner.
Boost the car's speed and we can see how the i60 handles it – it maintains even speed changes through this tight turn, like a smooth-as-jelly slide through the intestine.
If human driving is like using a scalpel with teeth and grain, then the experience i60 brings about feels more like stretching a rubber band, smooth and silky.
And finally, the "soulism" that embodies the high end of this system.
Here, i60 excels in two areas: one being features that we usually find uninteresting or underused, making them usable and effective.
For instance, the roaming feature: While the current WRD 3.0 from WenYuan Zhixin doesn't yet have parking lot route memorization, it is the first system we've experienced that can fully autonomously navigate out of a parking lot and into a parking space via the roaming function, providing us with a complete P2P experience.
On the other hand, this car has a kind of biologically intelligent spirituality in certain scenarios.
Past the tracks, ahead is pitch black, which is actually a dead end.
We were expecting to test the i60's off-road capability, but it turned out to be a joke. It refused to enter and instead performed an extreme three-point turn at a crossroad, clearing the bumpers of the traffic barrier to complete the recovery.
Quick and nimble.
But we can’t let up now. Let’s turn the car around and see how it handles the end.
It tried to turn around, but failed due to very narrow space, with obstacles in front and behind it.
At this point, some friends said: Can't you reverse out?
This time it wasn't activated.
Here's a success case.
It’s another dead end, this time at the hospital.
We followed the car in front into a dead end, and the car in front realized we were stuck and put it in reverse.
In that instant, the i60 didn't hesitate and started reversing until it reached the end of the road, then did a 3-point turn. The idea was sound.
i60 gets a good review for its parking assist system, which can recognize reverse lights and actively reverses with good adaptability in currently available systems that support reverse gear.
Turning right here leads into a closed compound.
As the black car didn’t seem to be leaving, the i60 started to push forward; just as the driver thought it was being a bit rude, it swerved to the right, making way.
Is it soully ? Then I smoothly entered the parking garage, handling side parking situations by waiting and passing, and handling parked cars on the side with ease, showing very low latency, like how a human would drive.
In the end, what exactly does this system excel at? If it's using L4's underlying algorithm to achieve L2, that's like a game-changer; and if it leverages Qualcomm Snapdragon 8650, which has only 200 TOPS of computing power, to bring this system to a car priced at 150,000 RMB or less, and still offers such superior ADAS experience, then that's like a one-button supercharging.
But we'd break this performance into two sentences: the upper limit of capability is very high, but the stability of the boundaries is still being worked on.
Three times it was all about rule boundaries and long-tail interpretations, not foundational capabilities failing; three places were not smooth, it wasn't 'not possible', but rather 'not decisive enough' or 'not leaving enough margin."
This accurately explains the personality of WRD 3.0: It's like an extremely efficient driver who is willing to make decisions proactively, skillfully merging, weaving through traffic, and finding space in tight situations; however, when rules and spatial semantics become ambiguous, it still needs more data to fill in the "should" and "shouldn't".
As a dark-horse in the 2026 ADAS (Advanced Driver Assistance Systems) world, WeRide has not yet launched many models, so it may be unfamiliar to most people.
But as a long-time player in L4, offering L2 assisted driving for passenger vehicles is something that has been a bit delayed but is now coming.
It's rumored that in addition to GAC and Chery, more vehicles featuring WeRide's solutions will be making their debut to the public next year.
Will WenWay V, Momenta, and Horizon be the three major players in the supplier tier? Stay tuned for the Intelligent Car Studio.
(Sources: Intelligent Car Studio - Shao Peiheng)