Rate limits and quota
Two limits apply: requests per minute, per key, and total calls per month, per account.
| Plan | Per minute (per key) | Per month (per account) |
|---|---|---|
| Free | 30 | 5,000 |
| Partner | 120 | 100,000 |
Successful responses carry your current position, so you never have to guess:
X-RateLimit-Limit: 30
X-RateLimit-Remaining: 27
X-Quota-Limit: 5000
X-Quota-Remaining: 4863
A 429 for the per-minute rate carries the rate headers and a
Retry-After, but not the quota headers — the quota is not read on a request that
was already refused.
What counts
- Successful calls and
5xx-free failures count against your quota. - Your own
4xxerrors do not. Getting a parameter wrong while you are building should not cost you anything. - Reference endpoints cost the same as chat, so cache them — they rarely change within a day.
The monthly quota resets at 00:00 UTC on the 1st. Rate limits are per key, so a runaway script cannot starve your production traffic if you use a separate key for it.
Need more?
Ask. These numbers are a starting point for people building something real, not a sales funnel — if you are doing something useful for the ummah and 5,000 calls is not enough, tell us what you are building and we will raise it.