Skip to content
FenixIAm by AIworks 2028 FenixIAm

Private and local AI

Your AI, in the place you choose.

The same Fenix, with its agents, its measurement and its permissions, in two different places: a European cloud just for your organisation, or your own server, without leaving the building.

Built on request

Two places for your Fenix

Private

A Fenix installation just for you, in Europe.

Nobody shares your Fenix installation.

We deploy your own Fenix in a European Azure region. OpenAI models are contracted in Azure’s EU data zone: your requests and their answers are processed within Microsoft’s EU Data Boundary.

Where it runs
A dedicated European Azure region
Models
The OpenAI models Azure offers in its EU data zone
Who maintains it
We do, with your keys
Who it is for
Those who want a dedicated installation with processing kept in Europe

Local

Fenix on your own server. Nothing crosses the door.

Your data does not leave the building.

As a tailor-made project, we install Fenix on your server with open models, OpenAI’s among them. Text agents work inside your network and you pay no provider per message.

Where it runs
On your server, inside your network
Models
OpenAI’s gpt-oss-120b and gpt-oss-20b, and other open models
Who maintains it
Your team, with our support
With voice
Voice agents need cloud models today
Who it is for
Those working with information that cannot leave the house

Why private or local

Control, by the way it is built.

  • Your data, under your control You decide where conversations, documents and reports are stored, and who sees them.
  • Your regulations, easier to meet GDPR and your sector’s rules apply to an installation you control.
  • A predictable cost Locally there is no cost per message: you pay for the machine and the upkeep, not the usage.
  • Works offline Locally, text agents keep working even if the internet goes down.
  • The same panel as always The same measurement, the same spend control and the same per-person permissions as in the cloud.
  • Voice, said plainly Voice agents (training, reception, languages) use cloud voice models today. Locally, Fenix works with text.

Models

OpenAI’s open models, on your machine.

Local · OpenAI gpt-oss-120b For production. Fits on a single 80 GB GPU.
Local · OpenAI gpt-oss-20b Lighter. Runs with 16 GB of memory.
Private · OpenAI EU data zone OpenAI models on Azure, processed within the EU Data Boundary.

Both OpenAI open models are licensed under Apache 2.0. Locally, capacity depends on the server: we recommend it for tasks where control of the information matters more than maximum power. The newest OpenAI models reach the EU data zone after the global deployment: in each project we confirm which ones are available. Each installation is built as a project, with your case in front of us.

Tell us which information cannot leave the house.

In a thirty-minute demo we look at your case, which level fits and what it would take to launch it.