Written by Cláudio Afonso | LinkedIn | X
Tesla’s Vice President of AI Software, Ashok Elluswamy, hinted Thursday at a significant update to the company’s Full Self-Driving (FSD) technology as the company gets closer to starting testing driverless ride-hailing services in California and Texas next year.
“V13 is looking 🔥,” Elluswamy wrote on the social media platform X. On October 31, he stated that the FSD team is “aiming to get to feature complete for unsupervised FSD with the v13 series.”
The v13 update will introduce technical improvements such as 36 Hz full-resolution AI4 video inputs, native AI4 inputs with redesigned neural network architectures, and a threefold increase in model size and context length. Tesla is also scaling its data inputs by 4.2 times and training compute by five times, supported by the Cortex training cluster. The update will include enhanced predictions for collision avoidance, adherence to traffic controls, and navigation.
Earlier this week, Tesla CEO Elon Musk responded to a customer’s review of the current FSD version 12.5.5.3. The customer said it “has handled just about everything imaginable.” Musk replied, “And V13 is trending to be ~500% better.” In a separate post, Musk said v13 “will be another 5 to 10 times better in miles between interventions,” referring to the frequency of human input required during autonomous operation.
Although a release date has not been announced, Musk said the update is “coming soon.”
Elluswamy joined Tesla in 2014 as a software engineer for Autopilot and has led the Autopilot software team since 2019.
Written by Cláudio Afonso | LinkedIn | X
