I’m experiencing an issue with the o3 model when using the new responses endpoint.
Even though I explicitly request detailed reasoning summaries with the following configuration:
"reasoning": {
"effort": "medium",
"summary": "detailed"
}
In most cases (more than 90% of the time), the model does not return any reasoning summary.
Here is an example of a typical response I receive:
{
"id": "resp_bGl0ZWxsbTpjdXN0b21fbGxtX...",
"created_at": 1751638574,
"error": null,
"incomplete_details": null,
"instructions": "...",
"metadata": {},
"model": "o3-2025-04-16",
"object": "response",
"output": [
{
"id": "rs_6867e2304ac481a...",
"summary": [], //<--empty reasoning summary
"type": "reasoning",
"encrypted_content": null,
"status": null
},
{
"arguments": {
"query": "..."
},
"call_id": "call_h1pEYvnkW...",
"name": "function_name",
"type": "function_call",
"id": "fc_6867e23de5e0...",
"status": "completed"
}
],
"parallel_tool_calls": true,
"temperature": 1,
"tool_choice": "auto",
"tools": [...],
"top_p": 1,
"max_output_tokens": null,
"previous_response_id": null,
"reasoning": {
"effort": "medium",
"summary": "detailed"
},
"status": "completed",
"text": {
"format": {
"type": "text"
}
},
"truncation": "disabled",
"usage": {
"input_tokens": 6661,
"input_tokens_details": {
"audio_tokens": null,
"cached_tokens": 0,
"text_tokens": null
},
"output_tokens": 560,
"output_tokens_details": {
"reasoning_tokens": 512,
"text_tokens": null
},
"total_tokens": 7221
},
"user": null
}
So far I haven’t found a consistent workaround. Is anyone else experiencing this behavior?
Have you found any solution, or is it a known issue?