FSD v14.3 rollout

Tesla is rolling out FSD Supervised v14.3 over‑the‑air, promising roughly 20% faster reaction times plus upgraded neural nets for rare scenarios, improved parking, and better handling of emergency vehicles. The company framed the update as a step forward for supervised autonomy, and it’s already appearing in owners’ OTA feeds. (x.com)

Tesla’s newest driver-assistance update is not a new sensor or a new car part. It is a software rewrite that Tesla says cuts the system’s reaction time by about 20%, and owners started seeing the over-the-air package on build 2026.2.9.6 this week. (tesla.com) (electrek.co) Tesla calls the feature Full Self-Driving Supervised, and the last word is the important one. Tesla’s own support page says the system can drive the car “almost anywhere” only under active driver supervision and does not make the vehicle fully autonomous. (tesla.com 1) (tesla.com 2) That means the software is doing the steering, speed control, lane changes, and turns while the human is still legally and practically the backup. Tesla says more than 8.7 billion miles have now been driven with Full Self-Driving Supervised engaged, which is the data pool it uses to train newer versions. (tesla.com 1) (tesla.com 2) The biggest change in version 14.3 sits deep in the code, not on the screen. Tesla says it rewrote the artificial intelligence compiler and runtime “from the ground up with MLIR,” which is a tool for turning trained models into software that runs faster on the car’s computer. (electrek.co) (notateslaapp.com) Tesla also says it upgraded the reinforcement learning stage of training, which is the part where the system improves by getting rewarded for better choices and penalized for worse ones. In plain terms, that is like practicing thousands of driving decisions in simulation until the software starts choosing the cleaner move more often. (electrek.co) (notateslaapp.com) Another change is the vision encoder, which is the software layer that turns camera pixels into a usable picture of the road. Tesla says version 14.3 improves that layer in rare and low-visibility situations, strengthens three-dimensional geometry understanding, and expands traffic-sign understanding. (electrek.co) (notateslaapp.com) Those under-the-hood changes show up in very specific behaviors. Tesla’s release notes say version 14.3 reduces unnecessary lane biasing, cuts minor tailgating behavior, makes parking-spot selection more decisive, and improves maneuvering once the car commits to a spot. (electrek.co) (evshift.com) Emergency vehicles got their own line in the update notes. Tesla says the system now better understands emergency lights and sirens, which matters because flashing-light scenes combine odd vehicle positions, blocked lanes, and fast-changing right-of-way rules in a few seconds. (electrek.co) (notateslaapp.com) The rollout is not hitting every Tesla at once. Multiple reports say version 14.3 is going first to Hardware 4 vehicles, including newer Model S, Model 3, Model X, Model Y, and Cybertruck cars, with early public testers seeing it first. (electrek.co) (teslarati.com) Tesla is pairing that rollout with a new trial push. Tesla’s support page says new and existing owners of Model S, Model 3, Model X, Model Y, and Cybertruck can get a complimentary trial of Full Self-Driving Supervised version 14, including newer features such as Speed Profiles and Arrival Options. (tesla.com 1) (tesla.com 2) So version 14.3 is less about adding a flashy button and more about tightening the loop between what the cameras see and how fast the car reacts. Tesla is selling that as a supervised-autonomy step, not a hands-off robot driver, and its own documentation still says the human has to watch the road the whole time. (tesla.com) (tesla.com)

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.