Billing

How a request is charged

Each request costs its input tokens times the model’s input price, plus its output tokens (the model’s hidden reasoning included) times its output price. Input tokens served from cache are charged at the lower cached price. Prices are per million tokens.

input       2,000 tokens × $0.125  / 1M = $0.00025
reasoning     300 tokens
output        500 tokens
              800 tokens × $0.25   / 1M = $0.0002
cached      8,000 tokens × $0.0165 / 1M = $0.000132
total                                   = $0.000582

The amount is taken from your balance when the request ends; the Unaligned format returns it as cost. A request is refused with a 402 before anything runs when the balance would not cover the most it could cost: its input (estimated at four bytes a token) plus all the output it asks for (max_tokens, or the model's maximum when it names none), together with the requests of yours still running. It is also refused once the balance has reached zero while it waited for a free slot. One that still costs more than what was left (the estimate is an estimate) takes the balance below zero, and your next top-up covers it.

Top-ups

• Accepted: Monero, Bitcoin, Litecoin, Ethereum, Solana, Tether, USD Coin.
• From $5, no maximum. The rate is locked when the invoice is created.
• Credited after on-chain confirmation. Sending more than asked? All of it is credited.
• Sending less than $5? You have 60 minutes to send the rest; otherwise the funds go back to the refund address you gave. Without one after 30 days, they can no longer be refunded.
• First top-up: +5%, up to $100.

From code: the Payments API makes and follows the same invoices.

Usage

Dashboard → Usage shows tokens and cost per day and per model. Nothing finer is kept: there is no record of individual requests.