June 22, 2026
Xiaomi says its YU7 GT became the first car to set an autonomous lap record at the Nürburgring Nordschleife, completing the 20.8-kilometer circuit in 10 minutes and 29.483 seconds with no one in the driver’s seat. The Nordschleife folds more than 150 corners into the layout the industry treats as the hardest handling test in the world, and the car read and drove all of it unaided.
The time is the part the headline tends to drop. In May the same YU7 GT, with a professional driver, set a production-SUV record on the same track at 7 minutes and 22.755 seconds, more than three minutes quicker. A system that laps the Nordschleife in 10:29 is doing something no car has done in public, and it is also lapping at roughly the pace of a capable amateur rather than a racer. Both readings hold.
A flying lap is a different machine problem than a city street. There is no traffic and no pedestrian to track, but the car has to brake, turn, and hold grip near the limit at speeds reaching 300 kilometers per hour, judging weight transfer and a shifting surface in fractions of a second. The YU7 GT runs Xiaomi’s dual-motor V8s EVO drivetrain, 738 kilowatts and a 2.92-second sprint to 100, so the hardware was never the constraint. The software finding the edge and staying on the safe side of it is the achievement.
What Micromouse says happens next
Robotics already has a model for what happens after a machine posts a first time. In Micromouse, the contest IEEE launched in the late 1970s, small robots solve a 16-by-16 maze against the clock. Early mice crawled, feeling along walls with a stop-and-turn shuffle. Then builders gave them a shortest-path algorithm called flood fill, let them cut long diagonals instead of square 90-degree turns, and bolted on fans that suck the robot to the floor for grip. The quickest now finish the maze in under four seconds. The point is the mechanism: once a fixed metric and a repeatable contest exist, machines optimize against them faster than intuition expects, because they iterate without a driver’s reaction time or nerve.
A racetrack is a maze of a different kind. Solving it means finding the fastest legal lap, the ideal line through every corner plus the braking points, regen, downforce, and acceleration that connect them, and each circuit is its own puzzle the way each Micromouse layout is. The setup shifts from car to car and track to track, exactly as it does for the mice. What changes is the cost of a wrong answer. A mouse that misreads a wall resets and runs again; a car that misreads Turn 1 at 250 kilometers per hour is wreckage, so the iteration runs slower and more cautiously, and tire grip puts a hard floor under the lap time that no software can argue away. Those are conditions on the problem, not a different problem. The curve still holds its shape: quick early gains as the software learns the line, then a long grind toward the physical limit the human at 7:22 already sits near.
A record only matters if a crowd shows up
The moment a second carmaker posts an autonomous Nordschleife time, there is a competition, and contests on a fixed track are how the industry has always pushed cars forward, now with the driver being software still in development rather than a person. The track becomes a benchmark for perception and control at the limit, the kind of test that feeds back into how the cars underneath are built and tuned. For Xiaomi, the youngest name on the board, planting the first flag is the whole play. The lap is a first. Whether it turns into a record worth chasing depends on who shows up to beat it.
