# Models, privacy and usage

You choose the model for each message from the picker in the message box. Each model is
shown with its maker's mark, its name, and up to two marks after the name.

## The models

| Model | Flag | Shield |
|---|---|---|
| Claude Haiku 5.5 | Europe | yes |
| GPT-6 Luna | Europe | yes |
| GLM 5.3 Flash | none | yes |
| DeepSeek V4.1 Flash | none | yes |

Claude Haiku 5.5 is chosen for you until you choose another.

## What the marks mean

Rest the pointer on a mark to read what it means. These are its exact words:

- **The flag of Europe:** "Processed in Europe: this model runs in a European data centre
  (EU, UK, Switzerland or Norway), so your messages and its answers are processed there."
- **The shield:** "Private: the provider does not train on your messages and does not keep
  them."

## Usage and your allowance

Your use of the models counts against an allowance. Theorem shows it as a share, never as
an amount of money.

- **Usage**, in the menu under your name, shows **This month** with a bar, the percent
  used, and the day it resets.
- Beside the model's name, the message box says how much of this month's allowance is
  left.
- When it is used up, the message box says so, and the day it starts over.
- If no limit is set for you, **Usage** says **No limit is set.**