Responses.create() hangs indefinitely with large payloads on GPT-5 models (works fine with same inputs in O3 and small inputs tests in GPT-5)

API Bugs
1

When calling the Responses API (client.responses.create()) with GPT-5 models and large input payloads (~35k tokens), the API call never returns. The OpenAI platform shows the request was processed successfully (visible in logs/dashboard) and a response (output) was correctly generated, but the Python SDK hangs indefinitely waiting for the response. The same code for the api call works perfectly with O3 models or with smaller inputs on GPT-5.

1 Like
2

iā€™m seeing the same issue with the Responses API that started yesterday. On background mode, i see that large inputs to Gpt5 / Gpt5mini / Gpt4.1 seem to be queued indefinitely and never transitions to in-progress.

GPT 4.1-mini is slightly better, and the job transitions to in-progress, but what used to take < 1min to run now takes more than 5mins.

Are there any performance issues on OpenAI side?