The most affordable in the range: an AI assistant for a small team, meeting transcription and light vision models, in a computer that fits on a shelf.

It is a finished, tested computer, not a box of parts. If you need a different drive, another enclosure or a dedicated graphics card, we adjust it and send you a proposal within 12 working hours.
An assistant that answers from your company documents, with no data leaving the office.
To build and test agents and assistants with open models, without paying per query.
On a boat, an aircraft or a vehicle: artificial intelligence without a connection, with low power draw and remote management.
Models that understand images and video for inspection, surveillance or counting. With a professional graphics card if there are many cameras.
Meetings, dictation and calls turned into text and summarised on the computer itself.
As an OEM computer: the same configuration in every unit, a cloned image and a serial number.
We can fit a professional graphics card to any of the three compacts and it stays compact: it moves to an enclosure of about 10 litres, with the PCIe riser and a larger power supply included in the price. The AI model can sit entirely on the graphics card or be shared with the unified memory.
| Graphics card | Memory | Added to the computer | Complete Compact AI 32 |
|---|---|---|---|
| NVIDIA RTX PRO 2000 Blackwell | 16 GB with ECC | + 2,005 € | 4,070 € |
| NVIDIA RTX PRO 4000 Blackwell | 24 GB with ECC | + 3,820 € | 5,885 € |
| NVIDIA RTX PRO 4500 Blackwell | 32 GB with ECC | + 5,760 € | 7,825 € |
Prices exclude VAT, with the standard 10-litre enclosure. If your installation needs a different enclosure, the final price goes in the proposal. We also fit GeForce RTX cards if you do not need a professional one.
The same computer, ready to work where there is no technician and no connection: a boat, an aircraft, a vehicle or a remote site.
The models run on the computer itself: artificial intelligence keeps working at sea, in flight or with no coverage.
KVM over IP: screen, keyboard and power control at a distance whenever there is a link, including BIOS access.
A low-power processor and a 450 W power supply: it draws a fraction of what a workstation with a dedicated graphics card needs.
12 or 24 V power, mounting, connectors and a custom enclosure. Quoted with each order.
| Model | Processor | Unified memory | AI models that fit | Price excl. VAT |
|---|---|---|---|---|
| Compact AI 32This model | Ryzen AI Max 385 | 32 GB | Up to about 19 GB | 2,065 € |
| Compact AI 64 | Ryzen AI Max+ 395 | 64 GB | Up to about 42 GB | 2,940 € |
| Compact AI 128 | Ryzen AI Max+ 395 | 128 GB | Up to about 80 GB, or several at once | 4,820 € |
As a reference, a 70-billion-parameter language model compressed to 4 bits takes about 42 GB. For many simultaneous users, a dedicated graphics card performs better.
The processor and the integrated graphics share the same memory, so the graphics can use up to 24 GB to load the model. In a computer with a dedicated graphics card, the model has to fit in the card’s own memory, usually 16 to 32 GB.
No. The memory is soldered to the board, which is why there are three models: 32, 64 and 128 GB. Storage can be expanded through a second M.2 slot, and graphics with a dedicated NVIDIA RTX PRO.
When there are many simultaneous users, several cameras running machine vision or software that requires CUDA. With the card, the computer moves to an enclosure of about 10 litres.
Yes. It draws little power, takes up little space and has remote management over IP so it can be serviced without going on board. The 12 or 24 V supply, the mounting and the enclosure are adapted to each installation and quoted with the order.
Two to three weeks, depending on board availability. It has a 3-year warranty, including continuous 24/7 use, and a single point of contact for the whole computer.
A proposal within 12 working hours
Tell us what it is for and we will reply with a configuration and its price. If you already have a quote from another brand, send it over and we will prepare the equivalent.