Local AI

Local AI on your own server: data stays in the company

The language model can run on a server in your office or server room. Documents, emails and recordings never leave, and the assistant keeps working without an internet connection.

Managed by us, no server of your own

No server room? We set aside a DGX Spark with a local model for you in our EU data centre and connect it securely to your systems. The hardware serves only you, we run it, and you pay a monthly fee.

NVIDIA DGX Spark

A powerful computer the size of a shoebox with 128 GB of memory. For offices and companies with up to about a hundred users. It fits on a desk or in a server room.

GPU servers

For larger organisations and hospitals. NVIDIA GPU servers, Kubernetes and high availability, connected to Active Directory and your monitoring.

Combined with the cloud

A local model handles sensitive work, while demanding tasks that involve no sensitive data can go to Claude or Codex. You set the rules.

Who local AI is for

Which models we use

Open models such as gpt-oss, Qwen, Llama, Mistral and DeepSeek. We pick the model for the task, the language and the hardware, and update it when a better version comes out. The model comes with search across your documents, speech recognition and connections to your business systems.

What we deliver

Hardware, installation, model selection and tuning, integration with your systems, roles and permissions, backups and monitoring. After launch we handle updates and keep an eye on operations.

Frequently asked questions

How large a model can DGX Spark run?

DGX Spark has 128 GB of unified memory and runs open models of up to roughly 200 billion parameters. Two linked units can run even larger models.

Does local AI need an internet connection?

No. The model, document search and speech recognition all run on a server in your network, and the system can be fully offline.

Are local models as good as cloud models?

For everyday business work, yes. For the hardest tasks, such as complex programming, cloud models are still ahead, which is why we also offer a combination.

Do we have to buy the hardware?

No. We can run a dedicated DGX Spark for you in our EU data centre and connect it to your systems. Or we deliver the hardware to your office, or lease it to you.

What does the hardware cost?

It depends on the number of users and the size of the model. A small office needs one DGX Spark, while large organisations get GPU servers. We include the price in our proposal.

Want to know how much time an assistant would save you?

Tell us in a few sentences what your team does. We will reply within one business day and set up a free consultation.