Pricing
Pay per token, in rupees
Top up a rupee balance and pay the published rate for whichever model you call. No subscription, no seat licence, no dollar card, and Rs 1,000 of credit to start.
LLM Gateway
Pay-as-you-Go
Top up your account and pay per token usage as per models per-million-token rate.
Included
- Access to every P/9-hosted model with one unified, OpenAI-compatible API
- Compare models on published pedigree, latency and price
- Multiple API keys for different projects, each with its own scope and cap
- Cost and performance monitoring, per key and per model
- Chat support (48h, working days)
- Unlimited seats
Enterprise Gateway Platform
Custom price
Quoted per organisation, for dedicated capacity and enterprise terms.
All the previous features +
- High rate limits and bulk discounts
- Private deployments and specific compliance needs
- Custom latency and cost optimisation
- Dedicated support, SLA and professional services
- Custom billing and invoicing structures
- Pay as you go
- No hidden costs
- No API call limit
Every model and its rate per million tokens in rupees. Working out a budget before you write any code: what a million tokens actually costs, and which limit produces which error.
Common questions
Pricing, answered.
How is a request priced?
Each model has two rates: one per million input tokens, one per million output tokens. A request is billed for the tokens in the prompt at the first and the tokens in the completion at the second, settled against your prepaid balance the moment it finishes. A million input tokens and a quarter of a million out on DeepSeek-OCR 2 is Rs 9.02 plus Rs 2.26, so Rs 11.28. Nothing is estimated and nothing is averaged over a month.
How much does an AI API cost in Pakistani rupees?
It depends on the model and on how many tokens a request uses. Rates are published per million tokens: DeepSeek-OCR 2 is Rs 9.02 per million input tokens and Rs 9.02 per million output tokens, and every route's rate is published on the models page. Output tokens cost more than input tokens on every model, which is why a chatty assistant costs more than a long document it only reads.
Is there a subscription or a minimum spend?
No. P/9 is prepaid and pay as you go: you top up a rupee balance and each request draws against it at the rate published for that model. There is no monthly fee, no seat licence and no minimum.
Do I need a dollar card or a foreign bank account?
No. The balance, the rates and the invoice are all in Pakistani rupees, so there is no card in a foreign currency and no exchange-rate movement between the day you top up and the day you spend it.
Is there a free tier?
Every new organisation starts with Rs 1,000 of credit, enough to run real traffic through several models before deciding. It is credit rather than cash: it pays for requests and is not withdrawable, which is also why it sits outside the refund policy.
What can I cap, and what happens at the ceiling?
Each API key carries a rupee cap, a model scope and an expiry, and the organisation carries a budget above all of its keys. The gateway enforces them on the request path rather than in a nightly job, so a key that reaches its ceiling stops there and a request that would take the balance below zero is refused instead of run. A budget that fails closed is the difference between a surprise and a bounced request.
Can I see what each individual request cost?
Yes. The request log records the model, the key, the token counts in and out, the latency and the settled cost of that call in rupees. A bill is a sum of lines you can read rather than a figure you have to trust.
What if I need reserved hardware or an invoice?
That is dedicated capacity: Ascend NPUs held for your organisation alone, quoted per engagement rather than drawn from prepaid credit, with contract terms to match. The dedicated page has the detail and the form.
More on how this works: what a sovereign AI gateway is, the model catalogue and its rupee prices, and what the PK Endpoint commits to.