Pricing

# Pay as you go for the gateway. Enterprise for the rest.

A gateway request is paid per token from a prepaid balance. There is no subscription to buy.

The gateway plan gates no model and no gateway feature. Every model, and routing, redaction of shaped identifiers, a response cache, a prompt registry, batches and traces, from the first request.

[Models and rates](https://vatan.one/models)

#### Pay as you go: the Gateway

Available now

No monthly fee

A card, a balance, and nothing to sign.

- Model tokens paid per token from a prepaid balance, at the rate the models page shows
- Nothing further is added: no platform fee, no seat and no subscription
- A prompt prefix the provider answers from its own cache is charged at the lower cached rate, on the same terms as every other rate. Your usage screen shows it as a token count and a saving
- An OpenAI-compatible endpoint, streaming, on your own API keys
- Routes that name an ordered list of models and fall through when one produces nothing, and stop the moment one answers
- Redaction of shaped identifiers before a request leaves the estate, per detector. It finds card numbers that satisfy Luhn, IBANs that satisfy mod-97, national insurance and social security numbers, email addresses and API keys. It cannot find names, addresses or anything written in prose, and no pattern-based redaction can
- A response cache, exact then semantic, off until you choose a mode
- A prompt registry with numbered versions called as name@3, where a published version cannot be edited
- An asynchronous batch queue for chat completions. It draws on the same balance, budget and rate limit as a live call. It runs at the provider's own batch rates, on the chat models that service accepts and prices below its live rate
- A per-request trace of every attempt: which model, what status, how long and what it cost
- Keys you create and revoke yourself, each with its own rate limit
- A prepaid balance that the meter draws down per request
- A spend cap on any key, monthly or for the life of the key. It is checked before the request and counted after it, so a key stops itself without touching the balance or any other key
- A list of source addresses on any key, so a key that leaks works from nowhere but your network
- Several people in one workspace, each an owner or a member, invited by email or admitted automatically from an email domain you have proved you own
- An audit log of who changed what: who invited or removed somebody, who issued or revoked a key, who changed a route or a prompt
- Usage and spend broken down by key, by model and by whoever issued the key, readable over the last 90 days
- Card top-up from $5, or a credit code
- Email support, best effort

[Create an account](https://vatan.one/signup)

#### Enterprise

On request

Get in touch

Everything beyond the gateway, switched on for your account and set up with you.

- Vatan Assistant
- Vatan Knowledge
- Vatan Agents
- Vatan Workflows
- Vatan Widgets
- Vatan Market
- Vatan GTM
- Vatan Meetings
- AI Opportunity Scan
- Agent Sprint
- AI Enablement
- Dedicated LLM
- Fixed-scope services quoted before work starts

[Request a demo](https://vatan.one/contact?reason=demo)

How the balance works

## Top up, then every request draws it down.

Nothing is billed afterwards. A request is metered per token and taken from the balance, and at zero the gateway stops instead of running up a bill.

**TOP UP**By card, from $5

**OR**Redeem a credit code

**PREPAID**Your balanceNo platform fee, seat, subscription or minimum

**EACH REQUEST**Is the balance above zero?

**AT ZERO**402, and it stopsNo bill runs up

**ABOVE ZERO**Metered per tokenAt the model's rate, as set out above

A prompt prefix the provider answers from its cache is charged at the lower cached rate.

Press enter or space to select a node. You can then use the arrow keys to move the node around. Press delete to remove it and escape to cancel.

Press enter or space to select an edge. You can then press delete to remove it or escape to cancel.

## Ask an AI to compare.

[Vatan vs in-house](https://vatan.one/compare/build-vs-buy)

[ChatGPT](https://chatgpt.com/?q=Compare%20Vatan%20(vatan.one)%20to%20building%20AI%20in-house%20for%20an%20enterprise.)[Claude](https://claude.ai/new?q=Compare%20Vatan%20(vatan.one)%20to%20building%20AI%20in-house%20for%20an%20enterprise.)[Perplexity](https://www.perplexity.ai/search?q=Compare%20Vatan%20(vatan.one)%20to%20building%20AI%20in-house%20for%20an%20enterprise.)[Google AI Mode](https://www.google.com/search?udm=50&q=Compare%20Vatan%20(vatan.one)%20to%20building%20AI%20in-house%20for%20an%20enterprise.)

Pricing questions

## Asked in every procurement cycle.

Where does our data go?A request goes to the provider serving the model you asked for, and the answer comes back. Every provider that can receive your content is named on /trust. The usage record keeps the model, token counts, latency, cost and status, and no prompt or completion text. Content is kept only when a feature needs it, such as the response cache, a batch, a saved prompt, a Knowledge collection or an agent run. The platform runs in one EU region, named with its host and country on /trust.

What does Vatan charge?For the gateway, a price per 1M tokens for each model, shown on the models page. It is drawn from a prepaid balance per request, with no platform fee, seat, subscription or minimum. Top up by card or credit code. At zero the gateway answers 402 and stops, and no bill runs up.

What happens when the model provider has an outage?A plain call retries once, about 250 milliseconds later, on the same provider. A route you define names several models across providers and moves on at a 429, a 5xx or an overload. It stops at what a retry only repeats: a bad request, a refused key, an unknown model. It also stops once a model produces output, because a completion bills and a second one charges twice.

What does Vatan add to the token price?Each rate is what that model costs us to buy, plus 5.00%. There is no platform fee, no seat, no subscription and no minimum.

Cached prompts, and when a rate changesWhen a provider answers part of your prompt from its own prefix cache, it charges a fraction of its input rate, as little as a tenth. Vatan charges you that cached rate on the same terms as every other rate. Your usage screen shows how many tokens were cached and what the difference came to. Where a provider publishes no cached rate the full input rate applies.

The price table is a time series. The meter reads the row in force when a request was made, so a charge from last week still reconciles against the invoice it produced. If the rate changes it changes here on the same day it changes in the meter, and a test fails until it does.

How do we pay?A prepaid balance, topped up by card from $5, or by redeeming a credit code. There is no invoicing, no purchase order and no net-30. Requests are refused with a 402 when the balance reaches zero rather than running up a bill.

Source: https://vatan.one/pricing
