Where the numbers come from

• Logged rides: fares and distances from ride receipts riders share with the site (one entry per physical ride, from public vehicles only). Each is a ride somebody actually paid for.
• Rider reports: app quotes, pickup waits and "no cars available" moments riders report below. They count only after a moderator reviews them.
• Published reports, used only where the site has too little data: Tesla's reported fare structure over time (historical; not assumed to be current), RobotaxiTracker's published 90-day medians, Obi's day-to-day price range, and reported Austin wait times. These estimates are marked "Modeled from published reports" and are always low confidence.

How the fare estimate works

A trip's typical fare is a base amount plus a rate per mile, fitted to the logged fares with a method that ignores extreme values, so one unusually expensive ride cannot move it. Recent rides count more: a ride 30 days old counts half as much as one from today, and fares older than 180 days are not used. Impossible values (for example $200 for a mile) are dropped. Cybercab and Model Y are priced separately once each has enough rides with a known vehicle type; until then they share one estimate, and no discount is assumed for either.

Quiet, Typical and Busy

These are Cybercab Hunter's own demand scenarios, not labels Tesla uses. Tesla's prices have become more responsive to demand, so the same trip can cost more at a busy time. Each scenario is fitted from the spread of real fares: Quiet sits at the cheaper quarter of observed fares, Typical at the middle, Busy toward the top (80th percentile). The range shown is what remains uncertain once the scenario is known. The scenario assumed for the current time comes from fares, quoted waits and "no cars available" reports at comparable times, and is used only if it predicted held-out rides better than always assuming Typical. Pickup waits get the same treatment.

Why the registered fleet is not the available fleet

A registration shows a car was registered, not that it is driving. A sighting or camera detection shows a car was seen, not that it was free to pick someone up. So the page keeps four numbers apart: cars documented by the site, different cars seen recently (each car counted once, however often it was seen), cars registered with the Texas DMV (statewide), and cars available to pick you up, which nobody outside Tesla can see. None of them is used as "cars waiting for you".

Confidence

High: many recent observations, and at least 70% of recent rides the model had not seen fell inside its range. Medium: a moderate number of reasonably recent observations. Low: few or old observations, or built from published reports. Insufficient data: no estimate is shown. The accuracy check under each city is recomputed every day on the newest observations, which are held out from the fit.

Updates

The estimates are recomputed once a day. Austin and Dallas are computed separately from their own data; Austin's prices or waits are never applied to Dallas.

Fleet-geometry check

The previous version of this page estimated waits from fleet size alone: the nearest free car is on average 0.5 ÷ √(free cars per square mile) away, times a road detour factor, over an average speed, plus a minute to dispatch. It ignores demand and queues, so it is no longer the headline estimate. For the selected city it gives —, from these inputs:

Cybercabs documented

Loading Live

Cars in service at a time

40% to 80% Assumed

In-service cars that are free

30% to 60% Assumed

Average speed to the pickup

18 to 28 mph Assumed

Published fare history

Historical context only; Tesla does not publish its current formula.