Licensed to be used in conjunction with basebox, only.
// admin
Default models
Overview
Which model a new conversation or an app uses when nobody chooses otherwise – and with what reasoning effort. Defaults apply per organisation and can be overridden per app. Set the defaults well and you rarely need to intervene afterwards.
What it is for
Most chats and apps run with the default values. With two settings you decide the answer quality, speed and token consumption of the whole organisation.
The defaults at a glance
| Default | Applies to | Set where | Overridable |
|---|---|---|---|
| Default model | All chats; all apps set to "Use default" | Model selection | Per app under Advanced fields → Model |
| Reasoning effort | All chats with a reasoning model (preset Medium) | Administrator, for the whole installation | Per app via Thinking effort (Low · Medium · High · Very high · Maximum); per chat by users (Low · Medium · High · off) |
| Temperature | All answers | Server default by the Platform Operator (AISRV_LLM_TEMPERATURE) |
Per app in the app editor |
Where to find the setting
- Default model: in your organisation's administration area – see Model selection.
- Model and thinking effort per app: "All apps" → Edit app → Advanced fields tab.
Step by step
Set the default model:
- Open the administration area.
- Choose the model that represents the right trade-off between quality, speed and cost for the organisation.
- Save. New conversations use this model from now on.
Let an app deviate from the default:
- Edit app → Advanced fields.
- Model: choose the desired model instead of "Use default".
- Thinking effort: set it to match the task – Low for rephrasing and translating, High or above for analyses.
- Save.
Back to default: set the Model field back to "Use default". The app then follows every future change of the default model automatically.
What happens on a change
- New conversations use the new default model immediately.
- Apps with a fixed model are not affected.
- Context size and answer style may change – inform your users when you switch the default model.
- The "About" page always shows the active model and its context size.
Notes
Recommendation for the defaults
- Default model: the model that is good enough for the most common use and economical at the same time. Special cases get their own app.
- Reasoning default: Medium. Users can switch up or down per chat; apps for analyses get a higher thinking effort.
- Temperature per app: low for knowledge-base apps and factual answers.
Note
- More reasoning = more tokens. A high default affects the consumption of the whole organisation; check the Dashboard.
- Not every model supports reasoning. With models without reasoning, the reasoning default has no effect.
- API requests choose model and parameters themselves per request; interface defaults do not apply there.
- On basebox Server, every selectable model has to be connected – see Enabling models.
Frequently asked questions
What happens if the default model is no longer available? Chats cannot answer until an available model is set as default. basebox does not silently switch to another model.
Can I set different defaults for chat and apps? There is one default model for the organisation. Apps override it individually via Advanced fields → Model.
Can users bypass the default? Only for the reasoning effort in the chat and in their own apps via the Model field – provided several models are enabled. They cannot bypass a restriction of a model to a group that way.
How do I find out which apps deviate from the default? Open the app in the app editor and check Advanced fields → Model. There is currently no aggregate view.
Need help? Contact support