Errors

Non-2xx responses return a JSON body in this shape:

429application/json— Rate limited

{
  "error": {
    "message": "Rate limit exceeded. Please retry after a brief wait.",
    "type": "rate_limit_exceeded",
    "param": null,
    "code": "rate_limited"
  }
}

HTTP status codes

HTTP	Retryable	Meaning
`400`	No	Malformed request — check `error.message` for details
`401`	No	Missing or invalid API key
`403`	No	Insufficient permissions or balance
`404`	No	Resource does not exist (e.g. unknown model)
`429`	Yes	Rate limit exceeded (default: 100 RPM)
`500`	No	Internal server error
`503`	Yes	Service temporarily unavailable

LLM-specific error codes

Error	Status	Description
`INVALID_API_KEY`	401	API key not provided or invalid
`MODEL_NOT_FOUND`	404	The requested model does not exist
`INVALID_REQUEST_BODY`	400	Request body format error
`RATE_LIMIT_EXCEEDED`	429	Too many requests — retry later
`TOKEN_LIMIT_EXCEEDED`	429	Token limit exceeded — retry later
`NOT_ENOUGH_BALANCE`	403	Insufficient account balance
`SERVICE_NOT_AVAILABLE`	503	Service temporarily unavailable

Rate limit

The default rate limit for all endpoints is 100 requests per minute (RPM).

When the limit is exceeded, the API returns HTTP 429 with a RATE_LIMIT_EXCEEDED error.

Error schema

error

messagestringrequired

Human-readable description of the failure.

typestring

Machine-readable error category.

codestring | null

Short code for programmatic handling.

paramstring | null

Name of the offending request field, when applicable.

HTTP status codes

LLM-specific error codes

Rate limit

Error schema

On this page