Docs · Production

Error codes

Errors use standard HTTP status codes. OpenAI-format endpoints return OpenAI-style error bodies; the Anthropic-format endpoint returns Anthropic-style ones.

{
  "error": {
    "message": "Invalid API key",
    "type": "invalid_request_error",
    "code": "invalid_api_key"
  }
}

{
  "type": "error",
  "error": { "type": "authentication_error", "message": "invalid x-api-key" }
}

Status	Meaning	What to do
`400`	Invalid parameters, malformed JSON, a parameter the model does not support, or input beyond the context length	Fix the request; do not retry
`401`	Missing, wrong or deleted API key	Check the auth header
`403`	Insufficient balance, or the key is disabled / not allowed to use this model	Check balance and key settings in the console
`404`	Wrong path or unknown model	Check the base URL (the OpenAI SDK needs `/v1`) and the model ID
`413`	Request body too large	Shrink images or trim context
`429`	Too many requests, or the upstream is rate-limiting	Back off and retry; lower concurrency
`500`	Internal error	Safe to retry
`502` / `503`	Upstream temporarily unavailable	Retry after a short delay
`504`	Upstream timed out	Use streaming or retry later

Requests that fail with an error are not billed. For retry strategy see Best practices.

Best practices

Rate limits