# Rate limits

> The per-key request limit, free model limits, spending limits and provider rate limits.

Clonvo doesn’t rate-limit paid models beyond an abuse ceiling on each API key. Free models have their own limits. Beyond that, what limits you in practice is your credit balance, any spending limits on your keys, and the providers’ own rate limits.

## Per API key

On every [plan](https://clonvo.com/docs/plans.md), each API key can make up to 10,000 requests per minute, across chat completions and the other endpoints that take a key. The ceiling is there to stop abuse, not to meter normal use. Over it, requests return `429` with the code `rate_limit_exceeded` and a `Retry-After: 60` header. Paid models have no other request-rate limit on Clonvo’s side: the provider’s limits are passed through (see below).

## Free models

Requests to `:free` models are also limited per organization, on every plan. On Free, these are the only models you can use.

| Limit | Value |
| --- | --- |
| Requests per minute | 20 |
| Requests per day | 50 |
| Requests per day, after buying $10 of credits in total | 1,000 |

Going over either limit returns `429` with a `Retry-After` header: the code is `free_model_rate_limited` for the per-minute limit and `free_model_daily_limit` for the daily one, which resets at 00:00 UTC. See [Free models](https://clonvo.com/docs/free-models.md#limits).

## Provider rate limits

Requests on Clonvo credits share Clonvo’s accounts with each provider. Requests on [your own keys](https://clonvo.com/docs/byok.md) count against your account’s limits with that provider.

When a provider rate-limits a request on your own key, Clonvo retries it on Clonvo credits, unless the key is set to **Always use this key**. When it happens on credits, Clonvo moves to the next [fallback model](https://clonvo.com/docs/fallbacks.md), if you listed any. If nothing else can serve the request, you get the provider’s `429`, without a `Retry-After` header.

## Spending limits

A key’s spending limit is a cap on dollars, not requests. Once a key reaches its limit for the period, its requests fail with `402` and the code `key_limit_exceeded` until the period resets. Spending limits are available from Pay as you go up. See [Authentication](https://clonvo.com/docs/authentication.md#spending-limits).

## Retrying well

-   Retry `429` and `5xx` responses with exponential backoff and jitter, starting around one second.
-   Respect the `Retry-After` header when one is present. Clonvo sets it on its own `429` responses: the per-key and free model limits.
-   Don’t retry `400`, `401`, `402`, `403` or `404`: the same request will fail again.
-   Add fallback models so a busy provider doesn’t block you (Pay as you go and up).