# Understand Qix usage limits

Learn the default message and usage limits, who is exempt, and when to retry a limited request.

Canonical: https://demo.qixcount.com/help/qix-ai/usage-limits

Who can do this: People using Qix with writable business access; free viewers cannot resend questions

You'll need: Qix enabled; non-exempt users must wait for their request or daily allowance to reset

Qix limits message size and checks request and daily text-processing usage before sending a question for an AI answer. Service settings can override the default usage figures.

## If a limit is reached

1. Read the retry message in the chat.
2. Allow the indicated wait to finish before resending.
3. Shorten a message if it exceeds the character limit.

The widget can retry a limited question automatically. Daily usage resets at the end of the application's day. The widget does not provide a usage dashboard, and the defaults in Good to know may differ from the active service settings. Contact support if the wait or allowance is unclear.

Some directly handled prompts and confirmations do not need an AI answer and are handled before the AI usage check. This does not remove the permission or confirmation requirements for actions.

Qix can make mistakes. Review its answers and any proposed action regardless of remaining allowance.

Next, see [Why Qix may not answer](https://demo.qixcount.com/help/qix-ai/why-qix-may-not-answer).

## Good to know

| Limit | Default behavior |
| --- | --- |
| Message length | 4,000 characters per message |
| Request allowance | 20 requests per minute when the selected service configuration has no specific allowance; fallback configuration uses 60 |
| Daily text allowance | 100,000 tokens per user in a business when the selected service configuration has no specific allowance; fallback configuration uses 300,000 |
| Owner exemption | Owners and users with all permissions bypass the request and token limits |

A token is a unit of text processed for a question or answer. It is not a message: longer discussions can use more tokens. Input and output count toward usage. Conversation summaries can also count.