Tesla FSD Just Crossed 12 Billion Miles, Adding Roughly 27 Million a Day

Tesla’s global fleet has crossed 12 billion cumulative miles on FSD (Supervised), a staggering amount of real-world driving for one assistance system.

The newest billion arrived in roughly 38 days. That works out to about 26 million miles a day, close enough to call the recent pace roughly 27 million.

The scale deserves attention, and so does the word “Supervised.” These miles came with drivers responsible for watching the road and taking over when necessary.

Tesla's global fleet has now surpassed 12 billion miles driven on FSD (Supervised). 🚗

• The milestone comes just one month after the fleet crossed 11 billion miles on June 9.
• Every mile adds to the real-world driving data used to further develop Tesla's Full Self-Driving… pic.twitter.com/z2qZfdr0yE

— Tesla Owners Silicon Valley (@teslaownersSV) July 18, 2026

Tesla publishes a dedicated safety page that explains how it collects the mileage and collision information behind its FSD reporting. Anonymous trip-mileage packets are sent after a vehicle shifts into Park, while separate event packets can be transmitted after a detected collision.

The company counts a collision in the FSD-engaged category when the system was active at any point during the five seconds before the event. That rule is designed to keep a last-second disengagement from removing the incident from the assisted-driving data.

Tesla also says the collision figures use a rolling 12-month window and are updated quarterly. The page does not assign fault, and its mileage totals should not be confused with a count of unique situations selected for neural-network training.

Those distinctions matter because 12 billion miles measures exposure and scale. It creates a vast pool of possible examples, while the usefulness of that pool still depends on what the system encountered, which clips Tesla retained and how effectively the models learned from them.

The jump from 11 billion miles on June 9 to more than 12 billion on July 17 puts the latest billion at about 26.3 million miles per day. A different crossing time within either date can move that average, which is why “roughly 27 million” is the honest way to frame it.

At that pace, the fleet would accumulate another billion supervised miles in less than six weeks. Few driving programs can gather feedback across that many vehicles, roads, weather patterns and driver decisions.

Volume alone cannot guarantee diversity. A million routine highway miles may be easier to collect than one rare construction-zone interaction that exposes a genuinely difficult weakness.

Tesla defines FSD (Supervised) as an advanced driver-assistance system capable of handling much of a trip under active supervision. It can follow a navigation route, change lanes, take forks, make turns, steer around vehicles and objects, and work through other common road situations.

The driver still has to remain attentive, keep hands ready and intervene whenever conditions demand it. Tesla’s own description draws a hard line between a highly capable assistance package and a vehicle that can accept full responsibility for the drive.

That boundary keeps the 12-billion-mile figure in the right category. It is a measure of supervised use at extraordinary scale, rather than 12 billion miles of driverless commercial service.

The milestone can still be enormously valuable. More use gives Tesla more chances to find repeatable failure patterns, compare software versions and see whether a change holds up outside a controlled test route.

The next challenge is turning that flood of experience into behavior that feels consistent to the person behind the wheel. A system can complete the same maneuver safely and still frustrate its driver by choosing the wrong lane too early, hesitating at the wrong moment or repeating a preference the driver keeps correcting.

Tesla may be preparing to attack that problem directly. In a separate exchange about FSD leaving a carpool lane, Elon Musk said the vehicle will begin remembering specific interventions and matching each person’s preferences.

The car will start to remember your specific interventions and match each person’s individual preferences

— Elon Musk (@elonmusk) July 18, 2026

Tesla makes the driver’s responsibilities explicit in the Model Y owner’s manual. FSD (Supervised) can require intervention at any moment, and the person in the driver’s seat must continuously monitor the vehicle, surrounding traffic and road conditions.

The manual warns that the system may fail to detect or react appropriately to vehicles, cyclists, pedestrians and other hazards. It also directs drivers to be especially cautious around sharp curves, poor visibility, construction, complex intersections and unfamiliar road layouts.

Personalization will have to operate inside those limits. Remembering that one driver prefers to remain in an express lane longer could make the ride feel smarter, provided the learned preference never overrides a safer or legally required choice.

Musk did not attach a version number or release date to the idea. Until Tesla ships and documents it, the statement points to product direction rather than a feature owners should expect in their cars on a specific day.

Twelve billion supervised miles gives Tesla a reservoir of experience no lab could reproduce. The harder test is whether the software can turn that reservoir into fewer repeated mistakes and more predictable decisions.

If Tesla can make the fleet learn at this scale while respecting each driver’s preferences, the next billion miles should feel different from the last billion. Drivers will notice the improvement long before another counter rolls over.