Reasoning_effort parameter not working with GPT-5.2 in playground

LangSmith Product Help Observability & Evals
1

I’m getting this error when trying to use reasoning_effort: medium with GPT-5.2:

TypeError("AsyncResponses.create() got an unexpected keyword argument 'reasoning_effort'")

The SDK doesn’t recognize this parameter even though GPT-5.3 should support it. I can’t use reasoning effort controls as a result. Is this a known issue?

2

hi @baptgr

I assume you are using Responses API (AsyncResponses). I so, try nested options:

API Parameter Format
Chat Completions API reasoning_effort="medium" (top-level)
Responses API reasoning={"effort": "medium"} (nested dict)

or try to bump up your langchain ecosystem packages. The latest version of langchain-openai does handle this conversion automatically for Responses API:

if "reasoning_effort" in payload and "reasoning" not in payload:
    payload["reasoning"] = {"effort": payload.pop("reasoning_effort")}

In langchain-openai, certain models automatically trigger the Responses API. Looking at the source code (base.py line 535-538):

def _model_prefers_responses_api(model_name: str | None) -> bool:
    if not model_name:
        return False
    return "gpt-5.2-pro" in model_name or "codex" in model_name

If you’re using gpt-5.2-pro** or any Codex model, the Responses API is selected automatically. Additionally, OpenAI itself may route newer models through the Responses API by default in the Playground.