AI requests are model calls served through the Raeto app and API, plus every routing decision: requests where Raeto picks the model and the app calls it with its own key, through the routing API or the free routing preview. Each counts once, on the UTC day it was served.

Tokens are input plus output tokens of model calls served through Raeto. Routing decisions add no tokens, since the partner runs the model.

Model usage names the six most-used models in the chosen period; everything else, and calls from before per-model records began, is grouped as Other. Models used counts every distinct model ever served.

Routing decisions count every time Raeto picked the model: Auto on app and API calls, the routing API, and the routing preview, split by the fast, mid or top model it chose.

Gaps. On a day when calls went unrecorded (an outage, or a logging gap between two normal days), that day is filled in from the nearest normal day, spread across models in that day’s proportions. The same figure appears on the operator dashboard.

Privacy. Everything here is aggregate. No wallet, key or app is ever named. Numbers refresh every minute; the latest day is still in progress.