Pricing

Per site or per GPU. Never per seat.

No token bill. Setup once, then a fixed running cost to run your model and keep it learning.

Proposed prices, not yet tested with buyers.

Start here

Test Lab

$4,900

for four weeks

One task, on your own data, in a sealed setup. We show whether an Eigna expert beats your baseline.

Credited against setup when you continue.

  • Not betterYou keep the report.
  • BetterThe test set becomes your first release gate.

Then

Setup once. A fixed running cost.

Proposed prices, not yet tested with buyers.

Eigna Local

Data never leaves.

Setup$3,250a site
Running$1,100a site a month
  • Up to 15 users
  • One expert scope
  • No token limit: the machine is the limit
Eigna Local

Eigna Hybrid

Data stays in the region you choose.

Setup$3,250a site
Running$1,900a site a month
  • Up to 15 users
  • One expert scope
  • 30 million pool tokens a month
Eigna Hybrid

Eigna Cloud

Data stays in the region you choose.

Setup$5,500a dedicated GPU
Running$7,700a dedicated GPU a month
  • Up to 50 users
  • Three expert scopes
  • 12-month term; $8,850 month to month
Eigna Cloud

An extra expert scope is $325 a month. Hybrid top-up: $170 per 30 million tokens, only if you opt in. Larger projects are scoped from the task brief.

The running fee

What the monthly fee pays for.

  • Running the modelOn your machines, in your region or on your GPU.
  • Keeping it learningTraining on approved work, tested at the gate.
  • Releases and rollbackEvery release on record, and a way back.
  • CareA named contact and a monthly service report.

Hardware. Run it on your own machines, or on an Eigna Box bought or leased. The first Eigna Box is planned.

When a flat fee wins

Real for heavy, steady use. Not for light use.

Against the cheapest tiers, renting stays cheaper, and we make no cost claim.

One Local site a month, proposed
$1,100
GPT-5.5 tokens a month at which that site breaks even, about
98M
Claude Sonnet 5.5 tokens a month at which it breaks even, about
275M

List prices in October 2026, three input tokens per output token, no caching. Whether one site serves that volume depends on the machine.

Questions and answers

Why not per seat?

Because success should not make your AI more expensive. You pay per site or per GPU, so a team that uses its model more pays the same.

Is the Test Lab credited?

Yes. The $4,900 is credited against setup when you continue.

Are these prices final?

No. They are proposed, in US dollars, and not yet tested with buyers. Your offer confirms them.

Test Lab

Start with one task and four weeks.

On your own data, in a sealed setup. If it does not beat your baseline, you keep the report.