Tesla's FSD v14 Lite software for older cars is here
What's the story
Tesla has started rolling out its FSD V14 Lite software to a select group of Hardware 3 owners. The update is designed to improve the driving experience of older Tesla models by incorporating advanced features from the newer ones. However, it still doesn't enable fully autonomous driving for the cars.
Update details
How the update works
The FSD V14 Lite update takes the driving behavior of the Hardware four version and applies it to both the cameras and compute of Hardware 3. This lets older computers learn how to handle scenarios using a newer stack as a guide. The process brings improvements from Hardware 4 to the legacy fleet, including reinforcement learning and offline models, according to Tesla's release notes.
User benefits
Improvements and new features in the update
The FSD V14 Lite update promises to improve both proactive and reactive responses across navigation, merges and forks, pedestrian interactions, traffic lights, and vehicle cut-ins. It also promises comfort gains like fewer false slowdowns, smoother steering, and consistent lane centering. New features include parking/unparking/reversing functions and Arrival Options that let drivers choose where their car parks.
Safety enhancements
Wider release will follow
Tesla's head of Autopilot, Ashok Elluswamy, had previewed the build as offering "basically have all the features that V14 for Hardware 4 has." He emphasized that the update is a compressed neural network rather than a reduced feature set. The most important change with this update is "significantly improved safety," Elluswamy said, adding that a wider release will follow once early-access feedback validates the build.
Hardware limitations
Limitations of the update
Despite the improvements, it's important to note that the FSD V14 Lite update doesn't enable unsupervised driving. This is a major limitation for the four million vehicles Tesla built with Hardware 3. Elon Musk had previously confirmed that these vehicles cannot achieve unsupervised Full Self-Driving due to memory bandwidth limitations of their older computer systems.