Tesla Pushes FSD v14 Lite Wide to Its Hardware 3 Fleet

AUSTIN, Texas — Tesla has begun the broad rollout of Full Self-Driving v14 "Lite" to its legacy Hardware 3 fleet, pushing software build 2026.20.6.11 to a widening pool of the roughly four million older vehicles on the road worldwide. It is the first time the v14 neural-network stack has reached HW3 cars at scale, and for owners who have watched newer AI4 vehicles pull ahead for the better part of a year, it is a long-awaited moment.

The wide release followed a public signal from Tesla AI chief Ashok Elluswamy, who identified build 2026.20.6.10 as the wide-release candidate before the team issued a quick bug-fix point release, 2026.20.6.11, and expanded it beyond the early-access group. The staged approach is classic Tesla: validate on a small cohort, patch, then open the floodgates.

A Long-Awaited Bridge for Older Cars

Hardware 3, the self-driving computer Tesla shipped for years, has less raw compute than the AI4 platform now in current cars. Many observers assumed the full v14 experience would never fit on the older chips. The "Lite" build is Tesla answering that doubt, delivering the core v14 planner and behaviors to HW3 owners rather than leaving them behind.

That commitment matters because it keeps a promise. Tesla has steadily grown paid autonomy on these cars, and FSD subscriptions have climbed sharply as adoption spreads. Extending v14 to the legacy fleet widens the base of vehicles feeding real-world miles back into the training loop.

What v14 Lite Brings

Owners on the new build report the smoother, more decisive planning that defined v14 on AI4 cars, along with refined Speed Profiles and better handling of complex intersections. The point release cleaned up early regressions, and Tesla has been shipping FSD updates every few weeks, a cadence that lets it iterate faster than any traditional automaker. Independent coverage from outlets like Electrek has tracked how each release sharpens parking, navigation and city-street behavior.

The practical upshot is that a 2019 or 2020 Tesla can now drive with much of the same intelligence as a car built this year. For a company that sells software as much as steel, that is a powerful demonstration of how over-the-air updates keep vehicles improving long after they leave the lot.

Scaling Autonomy Across the Fleet

The bigger picture is data. Every HW3 car running v14 Lite becomes another node in Tesla's fleet-learning network, surfacing edge cases and quietly improving the models that also power the robotaxi program. That program keeps posting eye-catching numbers, with Tesla's unsupervised fleet crossing hundreds of thousands of miles with a clean safety record.

By putting its newest driving intelligence on its oldest eligible cars, Tesla is compounding that advantage. Expect the rollout to keep widening in the coming weeks, and expect the miles-per-day flowing into Tesla's training stack to climb right along with it.