Billing & invoices

USD pay-as-you-go with zero model markup: how charges are computed, peak windows, cache discounts and reading your bill.

The formula

charge = non-cached input tokens × input rate + cache-hit input tokens × cache-read rate + cache-write tokens × cache-write rate + output tokens × output rate. Rates are USD per million tokens — see the full rate card. “auto” is billed at the model it routed to; routing itself is free.

What we charge

• Model calls: exactly the upstream provider's public rate. Zero markup.
• Top-up fee: 10%, and this is our only revenue. Example: pay $100, $90.00 lands in your balance.
• Failed requests are never charged and never touch your balance.
• No monthly fee, no minimum spend, no token bundles you're locked into.

Peak windows, caching and context tiers

• Where the upstream has peak/off-peak rates (e.g. deepseek-flash), we bill by the time the request actually happened. Peak is 2× off-peak.
• Peak window: 01:00–04:00 and 06:00–10:00 UTC, Monday to Friday. Every other hour is off-peak.
• Input tokens that hit the provider's prompt cache are billed at the cache rate (as low as 1/50 of standard). The saving is passed to you in full.
• Some models are tiered by the input length of a single request; rates rise above 128K.
• On models with a thinking mode, output is billed at the thinking rate — see the pricing page.

Balance and bills

• Signing up grants $1.00 of credit — enough to run the whole integration without adding a card.
• Balances are in USD and charges are exact to 1e-9 USD. The ledger stores integers, so there is no floating-point drift.
• Every line in the console's bill expands to: input / cache-read / cache-write / output cost, token breakdown, peak flag, request_id, latency and the price-table version used.
• Filter by model and export CSV — usable directly for internal cost allocation.
• Lines appear within 1–2 minutes of the call (upstream metering is written in batches).

Querying your balance

curl https://api.aigateway.ltd/key/info \
  -H "Authorization: Bearer ${AIGATEWAY_API_KEY}"
# returns: spend (USD, gateway guard value) / max_budget (USD cap) / models

For the authoritative balance and per-request detail, use the console (/console) — it is computed by the billing engine from provider rate cards. The spend field from /key/info is the gateway's coarse guard value and may differ slightly from your bill.