Rate limits

Three independent limits protect the platform and your budget. All of them apply to the whole account: every API key of the account shares them, and more keys never raise them.

Limit	`standard` plan	When exceeded
Requests	10 per second sustained (short bursts above it are allowed), 300 per minute	`429 rate_limit_exceeded`
Concurrent tasks (queued + processing)	20	`429 concurrency_limit_exceeded`
Credits held per UTC day	no cap	`429 daily_cap_exceeded`

Your current limits are in the plan object of GET /v1/account:

{
  "plan": {
    "name": "standard",
    "rps": 10,
    "rpm": 300,
    "max_concurrent_tasks": 20,
    "daily_credit_cap": null
  }
}

Need higher limits? Email support@fiveapi.ai with your account id and your expected volume.

Rate limit headers

Authenticated responses carry the state of the current minute window:

X-RateLimit-Limit: 300
X-RateLimit-Remaining: 287
X-RateLimit-Reset: 1792736640

Header	Description
`X-RateLimit-Limit`	Requests allowed per minute for your account, or for the key when it has a lower ceiling
`X-RateLimit-Remaining`	Requests left in the current minute
`X-RateLimit-Reset`	Unix time in seconds when the window resets

Every 429, and every 503 service_unavailable, also carries Retry-After: the number of seconds to wait before you retry.

HTTP/1.1 429 Too Many Requests
Retry-After: 3
X-RateLimit-Limit: 300
X-RateLimit-Remaining: 0
X-RateLimit-Reset: 1792736640

{"error":{"type":"rate_limit_error","code":"rate_limit_exceeded","message":"rate limit exceeded for this account (10 rps, 300 rpm)","request_id":"req_01M5KN37C3D4E5F6G7H8J9K0L1"}}

Concurrent tasks

The concurrency cap counts the tasks of your account that are queued or processing right now. A task stops counting as soon as it reaches completed, failed or cancelled. When you hit the cap:

wait for running tasks to finish (Retry-After suggests a delay), or
cancel tasks you no longer need. Cancelling frees the slot and releases the credits at once.

Polling

GET /v1/tasks/{task_id} counts against the request limit like any other call, so a tight polling loop is the easiest way to hit 429 rate_limit_exceeded. On the standard plan (300 requests per minute):

Polling interval per task	Tasks you can follow at once
every 3 seconds	15
every 5 seconds	25
every 10 seconds	50

Poll each task every 5–10 seconds and back off to 30 seconds for long generations; a 429 tells you to wait Retry-After seconds. To follow more tasks, call GET /v1/tasks?status=processing (one request returns up to 100 tasks) or use webhooks, which need no polling at all.

Stay within the limits

Poll gently. Polling counts like any other request: poll each task every 5–10 seconds with backoff (see Polling), or let webhooks tell you.
Queue on your side. Keep your own queue of pending jobs and submit only while queued + processing tasks are below plan.max_concurrent_tasks.
Honor Retry-After on every 429 and 503, and retry with the same Idempotency-Key.
Watch X-RateLimit-Remaining and slow down before it reaches zero.

Get started

Core concepts

Guides

Build with AI

Resources

Rate limit headers

Concurrent tasks

Polling

Stay within the limits

​Rate limit headers

​Concurrent tasks

​Polling

​Stay within the limits

Rate limit headers

Concurrent tasks

Polling

Stay within the limits