One law, every model.
Provider cost + 5%. That's the whole price list — the same cut the mini router charges on named models, and less than their +8% on auto. No per-model games, no bonus-token mechanics.
What a call costs.
Keys are billed at published rates from measured token usage; if the provider omits usage, an estimate is used. Direct wallet calls prepay an exact quote. The catalog includes our 5% markup; these are rates for models used in yesterday's ecosystem traffic:
| Model | Input / M | Output / M | A 1M-token chat (3:1) | vs mini |
|---|
"vs mini" is on the matched set, spend-weighted by yesterday's real ecosystem usage — see the full head-to-head.
How you pay.
Top up a key
Any amount — a tenth of a cent works — in USDG on Robinhood Chain or USDC on Base. Under the hood it's an x402 payment: your wallet signs one EIP-3009 message — no transaction, no gas, no Solana. The credit lands instantly and never expires.
Or pay per call
Agents with a wallet can skip the key: send X-PAYMENT on /v1/chat/completions. The quote includes our 5% markup, estimated input and the requested maximum output, with a $0.001 floor. The signed amount settles before inference.
Prepaid quotes
Direct wallet payments are not adjusted to actual usage, and unused output or a provider error after settlement is not automatically refunded. A retry can require a new payment. Use a topped-up key for measured token billing.
Receipts
Every settled payment returns an x-payment-response header with the on-chain transaction. The analytics page shows the same ledger live.