Error Handling

The OpenAI Responses API uses a consistent error format across all providers. This ensures that your integration remains robust even when switching between different underlying models.

Error Object

All errors return a JSON response with the following shape:

{
  "error": {
    "message": "The provided model is currently overloaded.",
    "type": "server_error",
    "param": null,
    "code": "model_overloaded"
  }
}

Common Response API Errors

Type	Code	Description
invalid_request_error	invalid_prompt	The input prompt contains content that violates safety policies.
invalid_request_error	invalid_api_key	The provided API key is invalid or expired.
rate_limit_error	rate_limit_exceeded	You have sent too many requests in a short period.
server_error	provider_error	The underlying model provider returned an unrecoverable error.

Handling Errors in Code

When using the OpenAI SDK, you can catch specific exception types:

from openai import OpenAI, APIError, RateLimitError

try:
    client.chat.completions.create(...)
except RateLimitError as e:
    print("Handle rate limiting (e.g. exponential backoff)")
except APIError as e:
    print(f"Handle API error: {e.message}")

For a full list of HTTP status codes and their meanings, refer to the General Errors Code page.