API Error Codes

The Infercom API uses standard HTTP response status codes to indicate whether an API request was successful or failed. When a request fails, the API responds with a JSON object containing details about the error.

The error envelope

Every error response uses the same shape. The error detail is nested inside an error object, and the response carries a request_id alongside it.

Example error response

{
  "error": {
    "message": "Invalid value for 'temperature': 99 is greater than the maximum of 2.",
    "type": "invalid_request_error",
    "param": "temperature",
    "code": "invalid_value"
  },
  "request_id": "dagkf1tcohtmhkppeq70"
}

Field	Description
`error.message`	A human-readable explanation of the problem.
`error.type`	The error class, for example `invalid_request_error` or `authentication_error`.
`error.param`	The request property that caused the error. `null` when no single property is at fault.
`error.code`	A stable machine-readable code. See Error code reference.
`request_id`	A unique identifier for the request.

Log request_id on every failed request. It is the field our support team needs to trace a request end to end. Include it in every report you open at support.infercom.ai.

/v1/messages uses the Anthropic error shape, not the envelope above. The body carries a top-level "type": "error", and error holds message and type only - there is no param or code. request_id is present on every status, though only some carry a req_ prefix - the 401 does not. See Anthropic SDK compatibility.

Three responses do not use this envelope, and one of them is not JSON at all. See Responses that do not use the envelope before you write your error handling.

Error categories

Use the following table to understand the various error categories.

HTTP status	Category	Description	Suggestion
400	Bad request	The server couldn’t process the request due to invalid syntax or input values.	Review the API request syntax and parameters.
401	Unauthorized	Authentication credentials are missing or invalid.	Ensure valid API key/token is included.
408	Request timeout	The server terminated the request due to excessive processing time.	Retry the request with optimized payload.
410	Gone	The requested model is no longer available (deprecated or removed).	Use a currently supported model.
422	Unprocessable entity	The request is well formed, but the model is not enabled on your plan.	Use a model listed for your subscription.
429	Too many requests	Rate limit exceeded for your subscription tier.	Implement rate limiting or upgrade tier.
429	Queue Full	Too many requests are currently in flight; the server declined the request.	Retry after a short delay or reduce concurrency.
500	Something went wrong	A server-side error occurred.	Retry later or contact support if persistent.
503	Maintenance	The model is temporarily unavailable due to maintenance.	Try again after some time.

If you are experiencing 500 or 503 errors, check the Infercom Status Page for any ongoing incidents or scheduled maintenance before contacting support.

Error code reference

Use the following table to understand and troubleshoot errors programmatically.

HTTP status	Error code	Description	Suggestion
400	`context_length_exceeded`	Input/output tokens exceed model context limits.	Reduce prompt and response length to stay within model limits.
	`invalid_type`	Parameter type mismatch (e.g., expected int but got float).	Check the expected parameter types and correct the input.
	`decimal_above_max_value`	Decimal value exceeds the allowed maximum.	Adjust the value to fall within the allowed range.
	`decimal_below_min_value`	Decimal value is below the allowed minimum.	Increase the value to meet the minimum required threshold.
	`integer_above_max_value`	Integer value exceeds the allowed maximum.	Provide a smaller integer within the valid range.
	`invalid_value`	Parameter value is outside the allowed range.	Check the accepted range for that parameter and resend.
	`unknown_parameter`	The endpoint does not accept that parameter.	Remove the parameter. Check the endpoint’s reference page for the parameters it accepts.
	`model_not_supported`	The model exists but does not serve this endpoint.	Use a model listed for that endpoint. Check Supported models.
	`unsupported_model`	The model is not served by `/v1/responses`.	Use `/v1/chat/completions`, or pick a model the Responses API serves.
	`model_not_found`	Provided model ID does not exist.	Verify the model ID and ensure it’s available in your subscription.
401	`invalid_authentication`	Invalid or expired API key.	Generate and use a valid API key from the Infercom Inference Service portal.
	`invalid_api_key`	The API key is incorrect or has been revoked.	Create a new key at cloud.infercom.ai and update your client.
408	`request_timeout`	Request timed out due to server load or latency.	Retry with a smaller or optimized request payload. Additionally, consider upgrading to a higher subscription tier to avoid this.
410	`model_deprecated`	The requested model has been deprecated and is no longer available. No endpoint returns this code today. `/v1/responses` returns `internal_error`, `/v1/messages` returns no `code` field, and `/v1/chat/completions` returns a plain-text body with no code at all. A fix is in progress with our platform provider.	Use a currently supported model listed in the documentation. Do not retry - a 410 is permanent.
422	`model_not_in_plan`	The model exists but is not enabled on your plan. Returned by `/v1/chat/completions` and `/v1/messages`. The other endpoints return `400` for the same condition.	Use a model listed for your subscription, or contact support to add the model.
429	`insufficient_quota`	Rate limit exceeded.	Upgrade your plan for higher quota.
429	`queue_full`	Too many in-flight requests; server declined the request.	Retry after a delay or reduce request frequency.
500	`internal_server_error`	An unexpected server-side error occurred.	Retry later. If the issue persists, contact support.
503	`maintenance`	Model is temporarily unavailable due to maintenance.	Try again after some time or check maintenance announcements.

Responses that do not use the envelope

Three paths do not use this envelope. On two of them the body is still JSON, but error is a string instead of an object. On the third the body is not JSON at all. All three omit request_id. Handle all three shapes before you read error.message.

Malformed image data

Returned by vision requests when the API cannot decode the image. HTTP 400.

{
  "error": "Unable to decode image: cannot identify image file <_io.BytesIO object at 0x...>",
  "error_code": null,
  "error_param": null,
  "error_type": "invalid_request_error"
}

A forced `tool_choice` that the model does not satisfy

Returned by function calling requests when you force a specific function and the model produces no valid call. HTTP 400.

{
  "error": "Specified tool_choice with function name \"book_flight\", but the model output contains zero valid function call.",
  "error_code": null,
  "error_model_output": "I can help you book a flight from Munich to Lisbon...",
  "error_type": "Invalid function calling output."
}

This body adds a field the envelope does not define, error_model_output. It carries the text the model produced instead of the tool call. Note also that error_type holds a sentence here, not a code such as invalid_request_error.

A retired model on `/v1/chat/completions`

Returned when the requested model has been deprecated and removed. HTTP 410. This response is not JSON. It is served as content-type: text/plain; charset=utf-8 with a plain sentence as the whole body:

The requested model (MiniMax-M2.5) is not available on SambaNova Cloud. For more information, please visit https://docs.sambanova.ai/cloud/docs/resources/deprecations. For further inquiries, please contact our support team at help@sambanova.ai.

Parsing this body as JSON raises a decode error. Check the response content-type, or catch the decode failure, before you read the body.

Official SDKs already survive this. The OpenAI Python SDK raises openai.APIStatusError with status_code 410 and the sentence in .body. The failure hits clients that call .json() on the raw response themselves - requests, for example, raises requests.exceptions.JSONDecodeError.

The same condition on /v1/responses and /v1/messages does return JSON.

Do not retry a 410. The condition is permanent. Change the model ID instead. See Deprecations for current replacements.

Defensive parsing

def read_error(response) -> tuple[str, str | None]:
    """Return (message, request_id) for all three response shapes."""
    try:
        body = response.json()
    except ValueError:
        # Not JSON. Some errors, such as a retired model on
        # /v1/chat/completions, return a plain-text body.
        return response.text.strip(), None

    err = body.get("error")
    if isinstance(err, str):
        return err, None
    return err.get("message", "unknown error"), body.get("request_id")

Overview

Inference API

Using the API

The error envelope

Error categories

Error code reference

Responses that do not use the envelope

Malformed image data

A forced `tool_choice` that the model does not satisfy

A retired model on `/v1/chat/completions`

Defensive parsing

​The error envelope

​Error categories

​Error code reference

​Responses that do not use the envelope

​Malformed image data

​A forced tool_choice that the model does not satisfy

​A retired model on /v1/chat/completions

​Defensive parsing

The error envelope

Error categories

Error code reference

Responses that do not use the envelope

Malformed image data

A forced `tool_choice` that the model does not satisfy

A retired model on `/v1/chat/completions`

Defensive parsing