[DigitalToday reporter Jinju Hong (홍진주)] Tesla's FSD V14 has improved significantly over the previous version on real roads, but a user review said it still showed limits in choosing parking locations, reversing and judging lanes at intersections. It has moved closer to handling a trip from departure to destination on its own, but critics said map and rear-recognition improvements are needed for unsupervised self-driving.

On Aug. 26 local time, electric vehicle outlet Cleantechnica reported that a user who drove a 2026 Model Y for about 3 months rated FSD V14 as significantly better than V12 after using it in northern Wisconsin and on interstate highways. The user described FSD V14 as similar to having a professional driver handle driving in most situations.

In actual driving, it steadily kept to the center of lanes, turned at intersections, and avoided pedestrians, bicycles, parked vehicles and construction vehicles. The review said it also responded appropriately when passing through roundabouts or stopping after detecting deer.

In particular, its ability to handle departure points and destinations improved significantly. The vehicle reversed out of a garage, left a gravel driveway and drove to the destination while managing signals, speed limits, passing and returning to its lane. At the destination, it selected a roadside spot, driveway or parking lot, entered the parking lot, found a space and backed into it.

The review said the parking process still left room for improvement. The user said it could not choose a parking space near the desired entrance at Walmart and did not recognise disabled parking signs. The user also said it sometimes missed the home's driveway on the return trip and did not support parking in a garage.

Its speed-control method also changed. Instead of the driver directly setting an exact speed, the system lets the driver choose a driving profile such as "Mad Max", "Hurry", "Standard", "Chill" and "Sloth". Standard drives about 5 miles per hour above the speed limit, Chill keeps to the limit, and Sloth drives more conservatively.

At intersections, improvements and limits appeared at the same time. It tended to come to a complete stop at stop signs and then wait about 1.5 seconds more, and the user said they slightly pressed the accelerator to reduce delays when they judged there was no cross traffic.

One of the biggest challenges was rear recognition. It could drive forward with a bicycle rack installed, but it did not support reversing. The user said the camera above the rear licence plate was partly blocked by the bicycle and rack, and the rear-facing cameras near the front quarter panels could not properly see the close area directly behind the vehicle.

Map accuracy issues also appeared. At an intersection in Iron Mountain, Michigan, the vehicle ignored a right-turn arrow marked on the road surface and went straight from the right lane. The user said that might be because the map did not sufficiently reflect information on turning lanes. The user also said a Costco in Wisconsin was still marked as vacant land on the map when the vehicle was delivered.

Overall, the review said FSD V14 showed a considerable level of self-driving performance across varied environments, from rural gravel roads to general roads and highways, but challenges remain before it can handle all situations without human intervention. How much it improves up-to-date map data and rear-recognition capability is expected to be a key variable in expanding unsupervised self-driving.