Tesla releases FSD V14.3.3 update
- Tesla began rolling out software version 2026.14.6.6 on May 17, adding Full Self-Driving (Supervised) V14.3.3 and bundling the company’s Spring 2026 update. (notateslaapp.com) - Tesla’s release notes said the update rewrote the AI compiler and runtime with MLIR, producing “20% faster reaction time” in FSD. (notateslaapp.com) - Teslascope showed the rollout on May 17 as pending on six AI4 vehicles, including Model S, Model 3, Model Y and Cybertruck. (teslascope.com)
Tesla began rolling out software version 2026.14.6.6 on May 17, according to Tesla release notes tracked by owner sites. The package adds Full Self-Driving (Supervised) V14.3.3 and brings Tesla’s Spring 2026 features to vehicles already running the V14.3 branch. Early tracking pages from Not a Tesla App, Tesla-Info and Teslascope showed the build appearing on a small number of vehicles on Sunday. (notateslaapp.com) Tesla’s published notes described the update as a point release focused on driver-assistance behavior, parking, edge-case object handling and a handful of new on-screen metrics. (notateslaapp.com) The company also kept its standard warning that Full Self-Driving is supervised and “does not make your vehicle autonomous.” (teslascope.com) ### Which software version is Tesla sending to owners? Version 2026.14.6.6 is the build now associated with FSD (Supervised) V14.3.3, according to Tesla release-note pages mirrored by Not a Tesla App and Tesla-Info. Both pages say the update adds Spring Update features to vehicles on FSD V14.3 and increments the FSD version to 14.3.3. (notateslaapp.com) May 17 is the release date listed on Teslascope’s tracking page for the build. That page said the software had been detected on six pending installs spanning Model S, Model 3, Model Y and Cybertruck vehicles equipped with Tesla’s AI4 hardware. (notateslaapp.com) ### What did Tesla say changed inside FSD V14.3.3? Tesla’s release notes said V14.3.3 upgrades the reinforcement-learning stage of FSD neural-network training and upgrades the vision encoder to improve performance in low-visibility and rare scenarios. The notes also said Tesla expanded traffic-sign understanding and strengthened 3D geometry understanding. (notateslaapp.com) The most specific performance claim in the notes was a “20% faster reaction time” after Tesla “rewrote the AI compiler and runtime from the ground up with MLIR.” The same notes said the update mitigates unnecessary lane biasing and minor tailgating behavior, improves parking-spot selection and maneuvering, and improves parking-location pin prediction on the map. (teslascope.com) ### Which edge cases does Tesla say it addressed? Tesla said the update improves responses to emergency vehicles, school buses, right-of-way violators and other rare vehicles. The notes also said the software improves handling of small animals and unusual objects that extend, hang or lean into the vehicle’s path. (notateslaapp.com) Another line in the release notes said V14.3.3 improves handling of temporary system degradations by maintaining control and automatically recovering without driver intervention, which Tesla said reduces unnecessary disengagements. Tesla also said the model is now unified between Actually Smart Summon, FSD and Robotaxi. (notateslaapp.com) ### What will drivers notice on screen and in Smart Summon? Actually Smart Summon’s maximum speed is now 8 mph, or 13 km/h, according to the release notes. Electrek separately reported that the new cap is up from 6 mph and applies in the V14.3.3 rollout to AI4 vehicles. (notateslaapp.com) Tesla also added a new intervention-free driving metric to the Self-Driving App. The notes said drivers can now see distance traveled in FSD without an intervention, and the app will show the driver’s longest intervention-free streak. (notateslaapp.com) ### Does Tesla describe the system as autonomous? Tesla’s own feature description in the release notes says Full Self-Driving (Supervised) can drive the vehicle “almost anywhere” under driver supervision, including lane changes, turns and parking. The same description says the system “does not make your vehicle autonomous” and tells drivers not to become complacent. (notateslaapp.com) The next visible milestone is broader installation data. Teslascope said on May 17 that the page would refresh until the first vehicle completed installation, while Not a Tesla App and Tesla-Info were tracking whether the build moved beyond an initial sub-0.1% share of cars. (notateslaapp.com) (teslascope.com) (tesla-info.com)