Settings › AI › ModelsFor administrators

Models

Decide where your team's prompts may be processed, which models are allowed and what new chats and images start with.

Under Settings › AI › Models you set the processing region, the allowed models and the defaults for new chats and images. Members choose in chat only from what you allow here. Every change applies at once and is recorded in the activity log.

Choosing the processing region

The region is the outer boundary: a model outside it disappears everywhere. Your data is stored in Switzerland regardless.

Click a region

Under Processing region, click one of the four cards. Each shows how many chat models can be used there.

The “Models” page with the “Processing region” Switzerland, European Union, EU + Switzerland (selected, recommended) and Global.

Confirm

The dialog says how many chat models are allowed afterwards, what new chats start with and whether security mode switches off. Click Change region.

RegionAllowed are models processing
Switzerlandin Switzerland
European Unionin Germany or in the EU
EU + Switzerland (Recommended)in Switzerland, in Germany or in the EU
Globalalso worldwide, with no commitment on data location

A new workspace starts with EU + Switzerland.

A narrower region can reset the default to the automatic choice and switch off security mode if its model lies outside. If members of a group that enforces security mode or Switzerland would have no suitable model left, Custodos does not save the region, and the dialog names the group. In the region Switzerland there is neither web search nor image generation.

Allowing or blocking models

Under Chat models each model has a card with the approximate credits per reply and a switch. A blocked model disappears at once from every member's model picker.

The “Chat models” section with the filter All, Switzerland, Germany and EU and model cards with a location mark, credits per exchange and a switch.
  • Out of region: the model has no switch. Its approval stays saved and applies again when you widen the region.
  • Awaiting deployment: the model cannot answer requests yet.
  • If you block the default model, first choose a New default model and confirm with Block.
  • The security mode model cannot be blocked. First choose another one under Setting up security mode.

Under Details on each card you find the data centre, context size, cost band (Economy, Standard, Premium) and residency (Enforced, Verified, Contractual). What residency means is explained in Where your data lives.

Who can use a model (groups)

With groups (on request) you choose Everyone, Selected groups or Nobody per model. Before saving you see how many members gain or lose access. Groups held to Switzerland or to security mode that therefore cannot use the model carry the mark Switzerland only or Security mode. The model new chats start with always stays available to everyone. If a model would become the choice of Automatic (recommended) that way, it cannot be given to some groups only.

Defaults for new chats and images

The “Defaults” card with the “New chats” menu open: “Automatic (recommended)” and the allowed models, with Claude Sonnet 5.5 selected.
  • New chats: with Automatic (recommended) Custodos chooses the model, or you choose an allowed one. In the members' model picker this entry is called Default.
  • Image generation: the model behind Create image. Without a choice, First allowed applies.

You allow image models further down in the section Image generation. If all of them are blocked, nobody can create images.

Best practices

  • Stay with EU + Switzerland unless an obligation requires a narrower region.
  • Allow a few clearly different models: an economical one for everyday tasks, a strong one for analysis.
  • Before narrowing the region, check which model security mode uses.

Frequently asked questions

Next steps

Share the page Choosing a model with your team.

On this page