Docs / Reference
Rate limits
600 requests per minute per key, reported in x-ratelimit headers.
Last updated
The limit
Each API key can make 600 requests per minute, measured over a sliding window. The Playground has its own limit of 600 RPM per account, separate from key limits. Budget packs have the same limit as pay-as-you-go. If you need more, contact support; a key's limit can be raised up to 2,000 requests per minute.
POST /v1/messages/count_tokens counts toward the limit too.
Headers
Every model response carries:
| Header | Meaning |
|---|---|
x-ratelimit-limit | The key's limit per minute |
x-ratelimit-remaining | Requests left in the current window |
x-ratelimit-reset | Seconds until the window resets |
Retry-After | On 429 and on retryable 503: seconds to wait |
Over the limit
The request returns HTTP 429 rate_limited and is not sent to a model:
{"error": {"message": "Rate limit of 600 requests per minute for this key was exceeded. Retry after 12 seconds.", "type": "rate_limit_error", "param": null, "code": "rate_limited"}}
Wait for the number of seconds in Retry-After before you retry.
Other 429 codes
A 429 can also mean a spending limit, not a rate:
| Code | Meaning | What to do |
|---|---|---|
key_limit_exceeded | The key reached its daily or monthly limit | Raise the key's limit, or wait until the time in the message |
project_budget_exceeded | The key's project reached its daily or monthly budget | Raise the project budget, or wait until the time in the message |
See 4xx errors.