Billing & Top-ups
How billing works
Xisu AI bills by token, with separate input and output prices and no minimum or monthly fee. When the charge happens, why failed requests are free, how interrupted streams are billed, and why console figures show a data-as-of time.
Billed by token, input and output separately
Cost of a call = input tokens × input price + output tokens × output price. Prices are per model; see the Models & pricing page.
No starting fee, no minimum spend, no monthly rent. If you don't call, you aren't charged.
When the deduction happens
In real time, as each call completes, out of your account balance. Failed requests — the ones that return an error code — are not billed.
If a streamed response drops midway, what was already generated is still billed: the gateway meters what it produced, not how much reached you.
Why pages say "data as of hh:mm:ss"
Balance, usage and billing are owned by the gateway. The console shows a snapshot together with the time it was taken. Reconcile against that timestamp.
If you refresh the console right after a call, you may need to wait for the next fetch before that call shows up. That's normal — nothing was lost.
No credit line, no post-payment
This is a product decision, not a missing feature.
Extending credit requires verified identity plus a way to recover the debt. This platform has no real-name verification, does not collect payment through card networks, and has no recourse on default — so whatever limit was granted would be the ceiling on bad debt.
Worried about "don't cut me off at 3am because the balance hit zero"? The console warns you before the balance runs out — keep a margin sized to your usage. See Top-ups and refunds.