Licensed to be used in conjunction with basebox, only.
// ai chat
Choose a model
Overview
Which AI model answers you is decided first of all by your organisation – and individual apps can deviate from that. This page explains how that choice comes about, where you can see which model is currently active, and what you can influence yourself.
Who decides on the model
Three levels work together:
- The organisation. Administrators choose the organisation's default model under Model selection. On basebox Server, the Platform Operator additionally determines which models are technically connected.
- The app. In the app editor, under Advanced fields, there is a Model field preset to "Use default". If the app's creator enters a specific model there, exactly that model answers in this app.
- You in the chat. You control how thoroughly the model works – via the reasoning effort – not which model answers.
Where you see which model is active
Open the user menu at the bottom left and choose "About". It shows the active model, its context size and whether speech recognition and the API are available. The context size is the number that tells you how much text a chat can hold – see Context window.
Inside an app: if a model of its own is set there, it deviates from the organisation default. When in doubt, ask the app's creator or your administrators.
What you can set yourself
- Reasoning effort – Low, Medium or High, or switch reasoning off entirely. This affects thoroughness, response time and token consumption. See Reasoning effort.
- In your own apps: the model. When you create an app, you can set a specific model under Advanced fields → Model, provided your organisation has enabled several. Leave it at "Use default" if the app does not need a specific model. See Create a new app.
- Personalisation – style and tone of the answers, independent of the model. See Personalisation.
Why organisations set the model centrally
Models differ in speed, cost, context size and capabilities – for instance whether they support reasoning or can call tools (connectors). On a server, the choice also depends on the available GPU equipment. Setting it centrally ensures that everyone in the organisation works with a reviewed, approved configuration.
Notes
Note
- Not every model supports reasoning. If the reasoning selector is missing from the input bar, the active model works without reasoning.
- A model change by your organisation can affect answer style and context size. New conversations use the new model.
- Via the API, developers choose the model per request from the models listed under
/v1/models. See OpenAI-compatible API.
Frequently asked questions
Can I switch between several models in the chat? The choice of model is made by your organisation or the app. What you control in the chat is the reasoning effort.
How do I know which model answers in an app? Under Advanced fields → Model in the app editor, if you are allowed to edit the app; otherwise ask the creator.
Why does the same model answer differently in an app than in the chat? Because the app has its own app instructions and possibly a knowledge base. The model is the same; the framing is not.
Can I use a model from the internet (e.g. GPT or Claude)? That is your organisation's decision. Which external model providers are an option per environment is described under Security & Compliance → Model providers.
Need help? Contact support