Tesla started sending out Full Self-Driving (Supervised) v14.3 with software version 2026.2.9.6. And this move puts a key part of the company’s autonomy plans in place. Elon Musk called it the “final piece of the puzzle” after a full rewrite of the AI setup.
AI system changes
The update rebuilds Tesla’s AI compiler and runtime on MLIR. That change cuts reaction times by 20% in every situation. A new vision encoder handles 3D geometry better and picks up traffic signs clearly even in bad light.
Reinforcement Learning now deals with hard cases like emergency vehicles or small animals at night. It manages tricky intersections and odd road blocks too. Drivers see less lane drifting and no more tailgating or slow parking choices.
How the rollout works
Hardware 4 cars get the update first and Tesla picked influencers to test it early. Employees ran beta checks before the wider push. Musk said the cars now act “sentient” in city traffic with sharp reasoning skills.
And Hardware 3 users wait for a v14 Lite version around mid-2026. Company filings point to this split based on chip power.
New features ahead
Pothole avoidance comes next along with reasoning that goes past just the destination. Eye tracking for driver checks improves as well. Talk points to Banish for self-parking after drop-off and Cybertruck Summon.
Fleet hazard sharing might link cars to warn each other. These steps build on v14.3 to make hands-free drives routine.
Tesla FSD v14.3: Official Release Notes
Full Self-Driving (Supervised) v14.3 includes:
- Improved parking location pin prediction, now shown on a map with a P icon.
- Increased decisiveness of parking spot selection and maneuvering.
- Rewrote the Al compiler and runtime from the ground up with MLIR, resulting in 20% faster reaction time and improving model iteration speed.
- Enhanced response to emergency vehicles, school buses, right-of-way violators, and other rare vehicles.
- Mitigated unnecessary lane biasing and minor tailgating behaviors.
- Improved handling of small animals by focusing RL training on harder examples and adding rewards for better proactive safety.
- Improved traffic light handling at complex intersections with compound lights, curved roads, and yellow light stopping – driven by training on hard RL examples sourced from the Tesla fleet.
- Upgraded the Reinforcement Learning (RL) stage of training the FSD neural network, resulting in improvements in a wide variety of driving scenarios.
- Upgraded the neural network vision encoder, improving understanding in rare and low-visibility scenarios, strengthening 3D geometry understanding, and expanding traffic sign understanding.
- Improved handling for rare and unusual objects extending, hanging, or leaning into the vehicle path by sourcing infrequent events from the fleet.
- Improved handling of temporary system degradations by maintaining control and automatically recovering without driver intervention, reducing unnecessary disengagements.
Upcoming Improvements:
- Expand reasoning to all behaviors beyond destination handling.
- Add pothole avoidance.
- Improve driver monitoring system sensitivity with better eye gaze tracking, eye wear handling, and higher accuracy in variable lighting conditions.
We’ve launched an official Tesla Owners Online Forum. Join early and claim your "Founding Fuel" badge.
