Tesla Rolls Out FSD v14.3: 20% Faster Response, Global Fleet Learning Enabled

Tesla began rolling out FSD (Supervised) v14.3 to early public beta testers in the U.S. on April 8, following completion of internal testing and validation by employees. This update is described as a foundational overhaul of the AI compiler and runtime environment, leveraging MLIR technology to improve vehicle response speed by 20%. The new version introduces networked interaction and inference capabilities between individual vehicles and Tesla’s global fleet, using rare or complex scenarios encountered by other vehicles—such as intersections with compound traffic signals, curves, and behaviors of small animals—to train neural networks. The vision encoder has been upgraded in parallel, enhancing recognition of 3D spatial structures and traffic signs, and improving perception of overhead or lane-intruding objects like low-hanging branches or construction equipment. System performance in low-visibility conditions has been optimized, potentially improving driving capability in adverse weather. The Cybertruck now receives FSD features equivalent to those on the Model Y and gains a new parking blind-spot warning feature to prevent collisions with approaching vehicles, cyclists, or pedestrians when passengers open doors. Although “Smart Summon” did not receive targeted optimizations and the “Autonomous Valet Parking” feature remains unavailable, v14.3 improves parking spot selection and decision-making efficiency. Available parking spaces are now displayed on the map with a dedicated “P” icon, and the system’s ability to automatically recover from temporary fallback states has been enhanced, eliminating the need for driver intervention. The release notes also preview upcoming enhancements, including the highly anticipated pothole avoidance feature. Tesla plans to extend AI inference capabilities to all driving scenarios and upgrade its driver monitoring system to improve accuracy under challenging lighting conditions, supporting eye tracking even when drivers wear sunglasses—ensuring effective attention monitoring during hands-free driving.