API Rate Limit
Also called Rate limit, Rate limiting, 429 error
A cap on how many requests or tokens you can send to an AI API in a given time, set by the provider and usually raised as you spend more.
What is an API rate limit?
A rate limit is the maximum amount of use an AI provider allows through its API in a period of time. Go over it and further requests are refused for a short while, usually with an HTTP 429 "Too many requests" error, until your allowance refills.
How does it work?
AI providers typically measure several things at once:
- Requests per minute: how many calls you make.
- Tokens per minute: how much text you send and receive, sometimes counted separately for input and output.
- Daily or monthly limits: caps on total use, or on spending.
Limits are usually set per organisation or project, and differ by model. Most providers raise them automatically as your account builds up spending history, in tiers. The error response often says how long to wait before trying again.
What is it commonly confused with?
- A spending limit. A rate limit controls speed; a budget cap controls cost. You can hit one without the other.
- An outage. A 429 means you're sending too much, not that the service is down. Overload errors from the provider look different.
Why does it matter?
Rate limits decide how many users an AI feature can serve at once, so check them before you launch. Good practice is to retry with increasing waits after a 429 (exponential backoff), queue large batch jobs, keep prompts short so they use fewer tokens, and use batch or caching features where the provider offers them. When you compare AI APIs, look at the limits on the tier you'll actually be on, not only the price.
