# Modes and costs

{/* AUTO-SYNCED SOURCE: this page lives in apps/app/src/modules/agent/docs/ and is mirrored into the docs app by `bun sync:module-docs`. Edit it in the module, not in apps/docs. */}

<Lead>
With the assistant you never pick an AI model, only a mode: Lite, Standard, or Premium. The mode determines quality and price, REOS AI takes care of the rest. Billing is fair, per message, based on actual cost, with bundling for small amounts.
</Lead>

## The three modes

<DefinitionList>
  <DefItem term="Lite (0.15x)">Fast and cheap, for simple tasks such as notes, translations, and quick questions. Lite is off by default and must be enabled by the workspace owner under **Settings > Assistant** in the **Modi** (modes) section.</DefItem>
  <DefItem term="Standard (1x)">The default and the recommendation for about 90 percent of requests: thorough, precise, with all tools and artifacts.</DefItem>
  <DefItem term="Premium (3.5x)">Highest quality with detailed reasoning, for negotiations, long listings, and other important answers.</DefItem>
</DefinitionList>

<Info title="The mode never changes answer length">
The mode only influences which model does the work and what the message costs. It changes neither the task nor the length or depth of the answer. Lite does not deliberately answer shorter, Premium not deliberately longer.
</Info>

### Switching mode in the middle of a chat

The mode sits at the top right next to the chat title and can be changed at any time, including mid-conversation. The change applies from the next answer onwards; answers already written stay as they are and are not recalculated. So you can run a piece of research in Standard and switch to Premium only for the wording that matters.

If the chat is already long, a note in the picker points out that the assistant will read the entire history again after the switch. The next answer therefore costs more once, after which it is back to the normal price of the mode. In short chats this is negligible, so the note does not appear there.

In a chat that someone shared with you, the mode is display only. Changing it is up to whoever owns the chat, because that is where the credits are charged.

## Model selection

REOS AI picks the concrete AI model automatically, matching the complexity of your request. You never have to deal with model names.

Owners and admins can still steer this: under **Settings > Assistant > AI models** (Einstellungen > Assistent > KI-Modelle) a fixed model can be pinned per mode, separately for text and image. Bringing your own provider keys (BYOK) is already listed there but still locked in the current version ("Bald verfügbar", coming soon).

## Per-message billing

Every message is billed based on the cost it actually incurred, multiplied by the mode factor (0.15x, 1x, or 3.5x). There is no flat price per message: a short question costs less than a long research task with many tool calls.

To keep your credit history readable, small amounts are collected: only once they add up to 5 credits is a single entry booked, visible as **KI-Assistent · gebündelte Nutzung (N Nachrichten)** (AI assistant, bundled usage, N messages). Each person only pays for their own usage, never for someone else's bundle.

## Cost displays

<DefinitionList>
  <DefItem term="Estimate pill">If your draft describes a foreseeably expensive bulk action, the composer shows an estimate up front, for example **≈ 1.4 Credits · Standard**.</DefItem>
  <DefItem term="Cost line">Below each answer you see what it cost.</DefItem>
  <DefItem term="Small-amount rule">Amounts below the small-amount threshold are hidden by default so you are not constantly reading cent amounts. You are still always charged correctly. The **Kleine Credit-Beträge anzeigen** (show small credit amounts) toggle lives under **Account > Appearance** (Konto > Erscheinungsbild).</DefItem>
</DefinitionList>

## Empty balance

When your credit balance is used up, the assistant does not start a new answer and reports: **Dein Credit-Guthaben ist aufgebraucht. Bitte lade Credits nach.** (your credit balance is used up, please top up credits). If your workspace uses extended usage with a budget, the message reads accordingly: **Dein Credit-Guthaben und dein Budget für zusätzliche Nutzung sind aufgebraucht. Bitte lade Credits nach oder erhöhe dein Budget.** How to top up is covered in the [top-up guide](/help/howto/topup-credits).

## Cancellation and rate limit

- **Cancellation:** If you stop a running answer, the part already written stays, and only the cost incurred up to that point is charged.
- **Rate limit:** Each person can start up to 20 chats per minute. Nobody reaches that in normal work; if you do, wait a moment and try again.

## Frequently asked questions

<Faq>
  <FaqItem q="Why do I not see costs on some answers?">
    Amounts below the small-amount threshold are hidden by default. Under **Account > Appearance** you can enable the display with **Kleine Credit-Beträge anzeigen** (show small credit amounts). This does not change the charge itself, which is always correct.
  </FaqItem>
  <FaqItem q="Why can I not select Lite?">
    Lite is disabled by default. The workspace owner enables the mode under **Settings > Assistant** in the **Modi** (modes) section.
  </FaqItem>
  <FaqItem q="Do I get longer answers in Premium?">
    No. Premium uses a stronger model for higher quality and care, but changes neither the task nor the answer length.
  </FaqItem>
  <FaqItem q="What does the entry 'KI-Assistent · gebündelte Nutzung' mean?">
    That is the collective booking of your small amounts: individual messages below the threshold are collected and booked as one entry from 5 credits, with the number of included messages in parentheses.
  </FaqItem>
</Faq>

## Further reading

Which tools can be part of a message's price is shown in the [capabilities reference](/tools/agent/faehigkeiten). Balance questions are answered in the [top-up guide](/help/howto/topup-credits). Back to the [overview](/tools/agent).