Structured output: JSON output is sometimes concatenated with itself (+ \n)

API Bugs
1

I’m experiencing a problem with structured output, in the ChatCompletion API. Concretely, the ‘content’ field, which is supposed to contain a JSON string, contains the same JSON string twice, separated by a newline character. Concretely, I get:

{
  "id": "chatcmpl-DigYwsHhd3XmY3QJKnk5XPf13JWi0",
  "object": "chat.completion",
  "created": 1779542414,
  "model": "gpt-5.4-2026-03-05",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "<JSON TEXT>\n<SAME JSON TEXT>"
,
        "refusal": null,
        "annotations": []
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 30501,
    "completion_tokens": 5718,
    "total_tokens": 36219,
    "prompt_tokens_details": {
      "cached_tokens": 3840,
      "audio_tokens": 0
    },
    "completion_tokens_details": {
      "reasoning_tokens": 1478,
      "audio_tokens": 0,
      "accepted_prediction_tokens": 0,
      "rejected_prediction_tokens": 0
    }
  },
  "service_tier": "default",
  "system_fingerprint": null
}

I hope OpenAI team can investigate the problem with the “id” field above.

We’ve added client-side logic to workaround this bug, but I thought it would be worth reporting. Observed only with GPT-5* models, not 4o, for instance.

2

To be noted : the number of “completion_tokens”, as returned by the API, corresponds to the full strings (with the content duplicated); so we’re actually paying twice for the output tokens…