MenuLimits & error codes

/

Limits & error codes

The API uses standard HTTP status codes. Error bodies follow your client’s format, so the OpenAI and Anthropic SDKs can read them as is.

Error format

OpenAI-format endpoints return an error object with message, type, and code. Anthropic-format endpoints return "type": "error" with an error object containing type and message. Errors that occur mid-stream are sent as an event with the same shape.

402

{
  "error": {
    "message": "GeraiPoints tidak cukup, silakan top up",
    "type": "insufficient_quota",
    "code": "insufficient_quota"
  }
}

The message text is written in Indonesian and may change. For program logic, check the HTTP status code and type/code, not the message text.

Status codes

Status	type / code	Meaning	What to do
400	`invalid_request_error`	Invalid request: the body isn’t JSON, `model` is empty, or `messages` is missing/empty. Also appears when the base URL is wrong so the path becomes `/v1/v1/…`.	Check the body and base URL.
401	`authentication_error`	The API key is missing, wrong, or revoked — or used on the wrong route (a bansos key on a paid endpoint, or vice versa).	Check the key and base URL — Authentication.
402	`insufficient_quota`	Not enough GeraiPoints for this request.	Top up GeraiPoints, then resend.
403	`permission_error`	The model isn’t in this key’s allowlist, or it is a plan-only model you don’t have yet.	Add the model to the key’s allowlist, or use another key.
404	`model_not_found`	Unknown model id, or the model isn’t currently open. Prefixed names such as `gp/…` or `openai/…` also return 404.	Copy the exact id from the model list.
413	`—`	The request body is too large for the server to accept.	Shorten the conversation history or reduce attachments.
429	`rate_limit_error`	Too many requests: the model is at full capacity, or the Unlimited/Bansos plan limit has been exceeded.	Wait as indicated by `Retry-After` (if present), then retry with a delay.
502	`api_error`	The model service is busy or having issues; the request is not billed.	Retry after a few seconds, or use another model.
503	`api_error`	The model is under maintenance or temporarily unavailable; the request is not billed.	Try again later or use another model.

Usage limits

Type	Limit
Paid key (gk-)	No fixed rate limit per key; usage is limited by your GeraiPoints balance. 429 only appears when the model is at full capacity.
Unlimited plan	Limits on requests per minute and concurrent requests per account. 429 responses include `Retry-After` and `x-ratelimit-*`. Details on the Unlimited page.
Free tier (Bansos)	Daily request quota and a per-minute limit. 429 responses include `Retry-After` and `x-ratelimit-*`. Remaining quota on the Bansos page.
Size & response	Bodies that are too large are rejected with 413. Response length (`max_tokens`) may be lowered automatically if your balance can’t cover the requested value.
API key	Up to 20 active keys per account; allowlist of up to 64 models per key.
Customer keys	Limits are enforced before a request is processed: a request that would exceed the remaining limit is rejected. One-time limit used up → `402 KEY_BATAS_HABIS` (raise the limit to add quota); daily/monthly limit reached → `429 KEY_BATAS_HARIAN` / `KEY_BATAS_BULANAN` with a `Retry-After` header. Paused key → `403 KEY_DIJEDA`. -unli and bansos models cannot be used with customer keys.

Retrying correctly

Retry only on 429, 502, and 503, with increasing delays (e.g. 1, 2, 4, 8 seconds) and a limit of 2–3 attempts.
Don’t retry 400, 401, 402, 403, 404 — fix the request first; the result will be the same.
Respect the Retry-After header when present.
Every retry is a new, billed request. If the cause is a timeout in your app, increase the timeout — see Agentic tool settings.