Choosing a model
Use "Default" or a specific model, compare credit usage and processing location, and learn why a model is missing or a different one answers.
An admin decides which AI models are available to you. You choose per chat and can switch at any time. The new model receives the whole conversation so far and the attached files.
Choose a model
Open the picker
Click the model picker at the bottom right of the message box. It shows Default or the name of the chosen model.
Compare models
Hover over a model. The info card next to it shows Credit usage, credits per answer, context size and processing location.

Pick the model
Click the model. It applies from your next message and stays chosen for this chat.
Default
Default sits at the top. The line below it names the model behind it: your own Default model from Personal settings, otherwise your group's default (on request), otherwise the workspace default. A new chat always starts on Default.
If nobody has set a default, Custodos may answer short questions in a short chat with a faster model. Questions with calculations, amounts or percentages, files, a brain, an agent or a connector stay on the full model. Under every answer you see the model that actually answered.
What the model info means
| Detail | Meaning |
|---|---|
| Credit usage | Low, Medium or High compared with other models. |
| Credits per answer | The typical cost of real answers over the last 30 days, or an estimate for a model with little use. |
| Token context | How much text the model reads at once: conversation, files and new message together. |
| Thinks before answering | The hidden thinking is billed too. Answers cost more than their length suggests. |
| "No web or brain search, and it creates no files." | The model uses no tools. It still reads attached files. |
| "Cannot read images." | Attached images stay out of the request. |
The code next to each model (CH, DE, EU) shows where it is processed. Which model runs where is explained in Where your data lives.
Why a model is missing or greyed out
- Your admin has not enabled it, or it runs outside your workspace's region.
- It is enabled for certain groups only (on request).
- It is greyed out because a brain is selected in the chat: models without tools cannot search a brain.
- It is temporarily unreachable: "This model is not available right now."
An admin manages what is enabled under Models.
When a different model answers
| Case | What you see |
|---|---|
| Your monthly budget, the workspace's credits or your group's budget are used up | A notice names the economy model you keep working on. The model picker is locked, and web search, Brain, files and images are paused. Nothing extra is invoiced. |
| The credits left do not cover the chosen model | The answer shows "Answered by …": a cheaper model answered. |
| The chat runs in security mode | The picker shows Fixed by security mode, see Security mode. |
How budgets work is explained in Budgets and limits.
Frequently asked questions
Your message and its files take up a large share of what the model can read at once. If the message alone is too long, it cannot be sent. Remove a file, shorten the message or choose a model with a larger context.
The economy model has a ceiling too. Once it is reached, chat is paused for the whole workspace until the next billing month starts or credits are added. Contact your admin.