You buy the hardware, you own it. You pay a setup fee once and a licence annually. Adding a user costs nothing, because the machine is yours. That is the whole model.
Setup + annual licence · support 20% incl. model updates · hardware at cost
Every tier is the same idea at a different scale: a box in your server room, set up on your corpus, measured before it is bought. Prices are "from" — the final quote follows the pilot's sizing, in writing.
One team, one corpus. Setup covers loading and indexing your documents, training for the team, and the pilot’s measurement protocol. The honest way to try sovereign AI.
The full appliance for a 50–250-person company. Setup includes corpus preparation, matching configuration, two training sessions, and a load test for your real number of users.
Heavier hardware class, more users, deeper integration with your systems, priority SLA. Architecture review and code escrow on request.
Hardware at cost +15–20%, bought in the EU under your name. The invoice is yours, the warranty is yours, the asset is on your books. We size the class to your corpus and your users during the pilot.
A small unit for one team and a focused corpus. Same software, same zero external calls — less machine.
A GX10-class box with large unified memory: it holds a serious open-weight model and your whole index in one unit, in your server room.
Workstation-grade GPU for more users and faster reading. The class the pilot's load test points to when the default is not enough.
Which class you need is not a guess — the pilot's load test decides it, on your real documents and your real number of users. How the box works
Licence is annual, paid upfront. Support is 20% and includes model updates. Hardware at cost +15–20%, bought in the EU under your name.
We are early, and we price that honestly: a discount on setup in exchange for an instrumented, public case study. Your pilot is measured against a written protocol — accuracy, hours saved, on your corpus — and the numbers are published with your name on them. You get a lower entry price; we get the proof we have not manufactured.
Before the pilot starts, we agree a reference sample of your real documents and an acceptance threshold — in writing. If the pilot does not pass the agreed threshold on your sample, we do not invoice the licence. The guarantee is the acceptance gate, not a promise on a slide.
Setup is one-time work: installing the box, loading and indexing your corpus, configuring matching, training your team, and running the load test. The licence is annual and pays for the software itself — the models, the index, and the answer engine running on your box.
Support is 20% and includes model updates. Open-weight models improve every few months; when a stronger one ships, we test it on your reference sample and swap it into your box. Support also covers fixes and help when your team needs it.
You do, from day one. It is bought in the EU under your name, at cost +15–20%, and sits on your books like any other server. It is not leased, and it is not ours.
The box, your corpus, and the index stay yours — they live inside your walls and we keep no copies. What stops is the licence and support: no more model updates, no re-tests, no help line. Nothing has to be given back, because nothing left.
The pilot is paid work, scoped and priced before it starts — a reference sample of your documents, an acceptance threshold agreed in writing, and a measured result. If the pilot misses the gate, you pay for the pilot work only. No licence invoice.
Because the box does not meter seats. It serves your team; the pilot’s load test proves it handles your real number of users. Adding a colleague costs nothing — that is the point of owning the machine.
Twenty minutes, no slides. Tell us about your documents and your team — we'll show the system live and scope the pilot that produces your quote.