The short version

  • A fixed annual licence: the price is agreed in writing before the year starts, so you know your budget in advance.
  • No tokens on your servers: with the licence we count neither tokens nor queries, so heavy use does not raise the bill.
  • A pilot first: you start with a paid pilot on your own documents, with pass criteria agreed before it begins. If you sign the licence within 90 days of the pilot, the pilot fee is credited in full against it.
  • The hardware is yours: we give you a short specification, and you buy the hardware from your own vendor.

What the price depends on

  • The size of the deployment: how many people will use the system and how many documents it covers decide the number of GPU nodes, and we size that with you in the Discover phase. How we size a deployment.
  • One organisation or a group: a licence for a group of companies differs from a licence for a single organisation.
  • Where it runs: on your premises, air-gapped, or in a private cloud in your country. If we run the infrastructure for you in a private cloud, a managed-service fee is added.
  • Integration work: connecting Orchestra to your internal systems is scoped and priced in the same quote, according to the number of systems.

What the licence includes

  • The Orchestra model itself, running on your hardware.
  • Search over your documents with sources cited in every answer, respecting who may see what.
  • The safety layer and the audit log.
  • The agent platform that connects the model to your systems and channels.

Hardware is not part of the licence. What comes with the model is described on the Orchestra page.

The hosted option: pay per token

If you are not ready to run Orchestra on your own servers yet, we host it for you and bill per token: USD 0.05 per million input tokens and USD 0.10 per million output tokens. In that case the model runs on our infrastructure, not yours, so it suits work whose data may leave your network. You can move to the fixed on-premises licence whenever your data needs to stay inside your network.

How to get a quote

  1. A 30-minute call: you tell us what the system should do, how many people will use it, and where the data must stay.
  2. Sizing: we agree with you the size of the deployment, the scope of the pilot and its pass criteria.
  3. A written, fixed quote: the pilot price, the annual licence price, and the hardware specification your vendor can price.

Book the call, or write to us.

Compared with paying per token

A cloud model's bill grows with every conversation, and nobody can predict how many tokens an organisation will consume in a year. A fixed licence does not move with use. Which one is cheaper for you depends on how much you use it: how to find the break-even point with your own numbers, and when the cloud is the better choice.

Frequently asked questions

Do you charge per token or per query?

Not on your servers. There, Orchestra comes with a fixed annual licence per deployment, and there is no meter on tokens or queries. Only the hosted option is billed per token.

Does the pilot fee count towards the licence?

Yes. If you sign the licence within 90 days of the end of the pilot, the pilot fee is credited in full against the licence price.

Why is there no public price for the licence?

Because the licence price depends on the size of the deployment, and that differs from one organisation to the next. After a short call and a sizing you get a written quote at a fixed price, before you commit to anything.

Does the price go up if we use it more?

Not while the deployment stays at the agreed size. If the deployment itself grows, with more GPU nodes or more companies in the group, we re-size the licence with you.

Who buys the hardware?

You do. We give you a short specification your own vendor can price, and we do not sell hardware. Orchestra runs on one GPU-class node rather than a cluster, which keeps the hardware cost down.

What about LeadX?

LeadX is a hosted service separate from Orchestra, priced on request.