“Premature close” when sending messages to Assistants API threads?

[dante3]

Hi everyone,
I’m running into a strange issue when working with the Assistants API and wanted to check if anyone else has experienced this or found a solution.

I’m sending a message to an existing thread using:

5:{"threadId":"thread_BZp7o5emhYRuGexN4j8N3JgJ","messageId":"msg_ujHzpWmArfLKpJN4wu90rwBF"}
3:"Premature close

The response I get back is:

Premature close

There’s no additional error message or context. The call just fails abruptly without returning a full response body.

[david_v]

I’m having the exact same problem and response status. It started last night (November 25, 2025) (GMT-5) and is still happening today (November 26, 2025). I still haven’t been able to figure out what’s going on.

[CHIOUKH_Lyes]

Same here – I’m seeing this issue both in our apps and directly in the Assistants Playground. For me, streaming cuts off as soon as a response takes longer than about 15 seconds, but anything faster than that finishes normally.​

I have tested this with several different models and get the same result each time, and this has only started happening in the last 48 hours or so.​

[OpenAI_Support]

Hi everyone, sorry about this disruption. We increased the timeout on the backend so you shouldn't see this error for streaming requests. It may take a few minutes for the fix to be fully deployed. Let us know how things look for you!

[gareth.smith]

Thank you,

I can confirm that we are now not seeing this error.
Havent tested everything but normal assistant messages, vector store and code interpreter work as before

[tirtho_cn3]

Hi,

The timeout increase has definitely helped — we’re now getting noticeably longer responses. However, the stream is still being reset abruptly with INTERNAL_ERROR on some runs.

You can reproduce the issue directly in the API Playground.
Example thread: thread_iUa3AJIcZ0uEs6vZSI351Oga

[Nicolas_Ippolito]

Even though timeout increase has helped, usually our assistants take more than 60 seconds (new timeout) to generate the data we need so we are still experiencing the same issue.

[caseyjenks]

60 seconds is still too low for our use case. Is there any chance we get get this bumped back up to what it was?

[OpenAI_Support]

Thanks all for the context - just shared with my team to review next week!

[OpenAI_Support]

This has been updated (as of ~Nov 28)! Please let me know if you're still running into any timeout errors and thank you all for the feedback.