Tesla Rolls Out FSD V14.3.12 With Faster Reaction Times and New Speed Profiles

Tesla has begun rolling out Full Self-Driving (Supervised) version 14.3.12, a significant software update that rewrites the AI compiler to improve reaction times and introduces new driver-adjustable speed profiles.

What Happened

The release notes for update 2026.38.5.5 detail a complete rewrite of the AI compiler and runtime using MLIR, which Tesla claims results in 20% faster reaction time and improved model iteration speed. The update also upgrades the neural network vision encoder to better handle rare and low-visibility scenarios, while strengthening 3D geometry and traffic sign understanding. On the user interface side, Tesla has introduced two new speed profiles: SLOTH, which offers lower speeds and more conservative lane selection than the previous CHILL setting, and MAD MAX, which allows for higher speeds than HURRY. Additionally, the system now supports 'Arrival Options,' allowing users to specify drop-off preferences such as Carpark, Street, or Driveway, with the navigation pin adjusting accordingly.

Why It Matters

For owners, the shift from precise speed offset adjustments to broader 'Speed Profiles' via the right scroll-wheel simplifies customization, though it requires users to adapt to new terminology and behaviors. The introduction of MLIR-based compilation suggests Tesla is optimizing its software stack for faster execution, potentially reducing latency in critical driving decisions. The enhanced driver monitoring system, with improved eye gaze tracking and variable lighting accuracy, aims to reduce false positives while maintaining safety oversight. Furthermore, the ability to view intervention-free streaks and distance traveled in the Self-Driving App provides users with more transparent data on their autonomous usage, a metric that has become increasingly relevant for insurance and regulatory discussions.

The Bottom Line

Tesla's FSD V14.3.12 update focuses on foundational performance improvements through compiler rewrites and expands user control over driving behavior with new speed profiles and arrival options, while also addressing specific edge cases like small animal detection and emergency vehicle response.