I’ve tried using the responses endpoint for gpt4.1, gpt4.1 nano, and a few hours ago some of the calls went through and some of them returned cutoff leaving the strict JSON broken (despite having a high max token limit). Now, for the past hour, the endpoint is no longer returning and is consistently timing out for every call.
Anyone else experiencing the same issue with the Responses API?
I know their Compliance API is down according to Open AI Status, but it doesn’t mention that anything is wrong with the Responses API. Coincidentally at the same time there was an update that they were working on a fix for the compliance API is when I started experiencing issues → unsure whether this fix that they are working on is affecting the Responses endpoint…
The API response seems to be extremely slow, especially when processing structured data like JSON to return structured output. I have no idea what’s causing it, and they haven’t even announced that there’s any instability. I don’t know how to explain to my customers why it’s this slow. It’s really unacceptable.
Same. I noticed that the responses API started returning logprobs as part of our streaming API responses starting last night. I am not sure if this has something to do with the performance issue, but same problem here. Frequent timeouts, and overall slow performance when it does respond.
Same here. Was trying to demo something and it just became several times slower than before. If it is not an outage, then it is an intentional trade off to reduce the performance per request in order to improve total throughput.
I have the same issue. The responses API is unusable slow for me (e.g. with gpt-4.1-mini). Similar problem I can see in the OpenAI Playground. Only in the Chat G pt Interface it’s fast.
Actually, I will try other providers when I’m less busy because the quality of OpenAI services are going down and down. However, right now, all the instructions, model settings, output structures, etc. are set for OpenAI’s models so I have to invest a lot time to switch my vendor.
For obvious reason I found this thread as well.
To convert PDF order file with ~60 products into json structured output it takes ~30 minutes using responses ( Playground ) and 2 minutes using Chat Completion ( Playground ).
It does work, it’s just freaking slow to use responses.
And all of that when they recommend using it for new projects…