A private LLM for your business.
A private LLM for business is a language model that runs where your organisation decides, under your own access controls, and is trained on your own work. Ertas trains one on your data for a defined task and runs it Ertas-managed, in your own cloud or on-premise. Your data stays private.
What is a private LLM?
A private LLM is a language model deployed for one organisation, with its requests, logs and training data kept under that organisation's control.
The term covers two separate choices: where the model runs, and what it was trained on. An open model you host yourself is private and still generic. A model trained on examples of your work also learns your formats, terminology and decisions.
How is a private LLM different from a public AI API?
A public AI API processes each request on the provider's servers, under the provider's terms. Those terms have improved: OpenAI's enterprise privacy page says it does not use business data to train its models by default, and may retain API inputs and outputs for up to 30 days to provide the services and identify abuse, as of October 2026.
A private LLM puts location, retention and access under your own policy, and adds training on your own work.
- Location
- You choose where the model runs and where its logs are kept.
- Retention
- Your policy sets how long requests and outputs are kept.
- Training
- Trained on examples your team approves, for the task you choose.
- Cost
- Per token, by seat or on a dedicated GPU. At volume, fewer tokens per task and a lower cost per token than the large frontier models.
Where can a private LLM run?
In three places, chosen by where your data is allowed to go. The model sits behind one API in each of them.
When the data has to stay in one country, run the model in your own cloud account in a region there, or on your own servers. Custom AI models in Australia works through one example.
- Ertas-managed
- We host, monitor and update the model. Your data stays private.
- Your cloud
- Your own AWS, Azure or Google Cloud account, under the controls you already have.
- On-premise
- Your own servers, inside your network.
What should you ask a private LLM provider?
These five questions cover most of what a security review will ask. Our proposed trust framework covers the measurement and change questions with a task-specific evaluation, a safety floor and a model record.
- Where it runs
- Ask for the region and provider of every environment, including logs and backups.
- Who can see it
- Ask which staff and subprocessors can access prompts, outputs and logs.
- How it is measured
- Ask for a test set built from your examples, and a score for every version.
- Who owns it
- Ask what you can take with you, and on what terms.
- How it changes
- Ask how a new version is approved, and what stays in service if it falls short.
Common questions.
Is a private LLM more expensive than a public API?
It depends on volume. For a task that repeats at volume, costs come down on two fronts: a lower price per token than the large frontier models, and fewer tokens per task, since the instructions and context repeated in every prompt are trained into the model. For occasional, open-ended work a public API usually costs less. Running costs are per token for variable or growing traffic, by seat (a flat monthly fee for unlimited tokens, with a concurrency cap) for steady traffic, or on a dedicated GPU for the highest volumes.
Can a private LLM use our documents?
Yes. Training teaches the model your task and formats, and retrieval gives it documents that change often. Many builds use both.
Do we own a private LLM that Ertas trains?
It is trained on your work and shaped by your team, and it runs on Ertas by default. Ownership and handover of the model are available with a minimum engagement commitment.
How long does a private LLM take to set up?
A free audit and data preparation come first, then a proof of concept of three to four weeks. The production model follows once your team approves the result.