Local language models
AI on your own hardware.
Open language models can run on your own machine or server. We assess which model fits your actual tasks and hardware, set it up and keep it running.
What we do
-
Choosing the model and hardware
We go through the work the model is needed for and suggest an open model that fits it. That also shows what kind of machine is enough.
- Walkthrough of the use case
- Model options and their limits
- Hardware sizing: GPU, memory, disk
- An assessment of which tasks suit local use and which don’t
-
Setup and deployment
We install the model on your machine or server and make it usable in everyday work.
- Installation on your own machine or server
- An interface for your staff
- An API for your own applications
- Model updates and swaps
-
Tailoring and support
We tune the model to match your work and continue supporting it after deployment, to the extent agreed.
- Instructions and ready-made prompts
- Your own documents available to the model
- Onboarding for users
- Maintenance and support as agreed
What a local model does
Fit is assessed against your tasks and hardware.
Local models come in many sizes. A suitable one handles a bounded use case well, for example questions and summaries about material you don’t want to send to an outside service. What works locally for you, and what is better done another way, becomes clear once we go through the tasks and the hardware you have. Cloud models such as Claude or ChatGPT do not run locally; they are connected over their APIs.
When the processing is configured to stay fully local, the material stays on your devices. If cloud integrations are included, selected data is transferred out through them. Those integrations are always agreed separately, and we document what moves. Data protection doesn’t end there: it also depends on how your machines, access rights and backups are handled. We go through those as part of the deployment. Regulated and classified environments are always assessed separately against the requirements that apply to them.
When data must stay inside your organisation
- HR: internal questions and summaries from salary and personnel records
- Finance: working through unpublished forecasts, cash flow and M&A calculations
- Legal: searching and summarising confidential contracts and case material
- Healthcare: patient record summaries for a professional to review
- R&D: searching your own technical documents and trade secrets
- Defence: handling military operational or technical documents in an isolated, approved environment
Decide before installing
- Which work the model is used for first.
- Which documents the model may use.
- Whether a suitable machine exists or needs buying.
- Who maintains and updates the model.
Tell us the work and the machine.
The package and price depend on the hardware, the number of users and how much support you need. A few sentences are enough to start, and we’ll tell you what is possible and what it costs.