Pricing and billing

Pay as you go

There is no subscription. Each model has a price per million input tokens and per million output tokens, in PKR, shown in the catalogue. A request costs prompt_tokens × input rate + completion_tokens × output rate, and that amount is settled against your organisation's prepaid credit when the request completes.

New organisations created through Continue with Google start with Rs 1,000 of credits, enough to integrate and evaluate before topping up.

Budgets and caps

Your organisation has one budget, in rupees, and every key and project spends against it. A key can additionally carry a credit cap of its own, a lifetime rupee ceiling for that key alone (see Authentication), and the caps of all your keys added together cannot exceed the organisation's budget.

Topping up is handled through the dashboard's Billing page, not through this API.

What happens at zero

The gateway checks the organisation's budget and the key's cap before dispatching a request. If either is exhausted the request answers 402 and nothing is sent upstream, so a run-away loop cannot take you into debt: budget exhaustion fails closed, never on credit. Top up and the very next request goes through; there is no delay while a cache expires.

Seeing what you spend

• Request logs in the dashboard show the settled cost of every request, in rupees.
• Per key and per model spend is on the Overview and on each key's detail panel.
• In your code, send X-K2I-Debug-Metrics: true on a streaming request and the gateway appends the cost of that request (in micro-PKR) after the stream; see Streaming and debug metrics.

Rates and currency

Prices are set in PKR and billed in PKR. i2 buys upstream capacity in dollars and carries the exchange rate on its side; the rate you see in the catalogue is the rate you pay, and it changes only when the catalogue does.

Streaming and token counts

Streaming and synchronous requests are billed on the same per-token rates. The authoritative figure for any single request is the settled cost on its row in the request log, which is what your balance is drawn against.

Dedicated capacity

If you need reserved hardware, an SLA or invoicing rather than prepaid credit, that is a conversation: see dedicated inference.