[Critical] Over 25% Assistant API Request Timeout Randomly

I hope this bug will be fixed soon.
I have misunderstood the situation.

My app is completely down due to this issue.
const evaluateTrend = async (content: string) => {
const assistant = new OpenAIAssistantRunnable({
assistantId: process.env.EVALUATE_ID as string,

const assistantResponse: any = await assistant.invoke({

return assistantResponse[0].content[0].text.value;

the day circled on the left is an average day, the day circled on the right is a result of the standard 3.5 turbo model hanging and continuing to charge me.

My hope is to create an automated system, but with outages like this, I may need to step away from openAI

It’s same for me. Suddenly the bot stopped working and requests are being timed out. Looks like someone deployed on friday and left for the weekend.

I have created an assistant in the openAI palyground, and now I am trying to retrieve the messages responses while using the assistant in my http request.
But I am unable to get the messages, I have followed the instructions mentioned in the documentation but I am still struggling to get the response.

Here are my request curls:

curl --location …/v1/assistants//messages’
–header ‘Content-Type: application/json’
–header ‘Openai-Organization: XXXXXXXXXX’
–header ‘OpenAI-Beta: assistants=v1’
–header ‘Authorization: Bearer XXXXXXXXX’
–data ‘{
“messages”: [
“role”: “user”,
“content”: “I am traveling from Doha to London on 23 of March 2024, give me flight booking suggestions”

I have also tried to create thread, then create message and then tried to run assistant api request as mentioned in the above shared documentation but I am just getting echo response.

curl --location …/v1/threads/<thread_id>/messages’
–header ‘Content-Type: application/json’
–header ‘Openai-Organization: XXXXXXXXX’
–header ‘OpenAI-Beta: assistants=v1’
–header ‘Authorization: Bearer <API_KEY>’
–data ‘{
“role”: “user”,
“content”: “I am travelling to Doha to London on 23 of March 2024, give me flight booking suggestions”

1 Like

Assistants seems to be completely broken at the moment, although no updates from OpenAI. I had to switch everything to use Chat Completions.


It is broken today. Bad responses, too much time. I asked at support and I was looking at incidents dashboard. No responses !

1 Like

I have the similar issue with Assistan API today. My requests get stuck on status queued

we did it fam


Since Frebruary, 27 API of ChatGPT-4 is too much slow and return errors. A simple document processing which took 2 minutes to process is now waiting for 10, 15 minutes. There is no incident in the incident or report from OpenAI. Anyone is perceving this behavior ? When OpenAI will fix it and create transparent process for incidents?


The incident is being investigated by OpenAI.
Here’s the latest update:


We’re seeing the same issues.

Even just creating a thread or checking on thread status is timing out (28 seconds). It fluctuates from a 1 in 10 calls issue to an almost every call issue…
Even on the calls that succeed they’re taking significantly longer (2-4 times) than even a couple weeks ago.

Really hoping the OpenAI team can sort this out before next Tuesday -
We’ve got a day of live demos booked!

I suppose they did warn us that it’s a beta product…

1 Like

Yeah, working with bleeding edge tech is scary sometimes… It’s come so far since 2019…

Same here! For more than a day now! Basically all calls via python SDK either take extra extra long, or literally never ever EVER return a reply. I implemented some retrying, and that helps for some, since often if a GET is just hanging, like to retrieve an assistant object for instance, usually just trying 1-4 more times real quick will hit right on one of them. But kinda less feasible for POST, cause it is unclear what is and isn’t updated when it just hangs, like when adding messages to a thread, or updating an assistant for instance… cause like i would want to double check first before a retrying to make some change liek that, but that would just be more requests to wait on… ugh.

Yes… it is a beta but c’mon!!! They should at least be on top of the outage status.

This forum is literally the only psychologically safe space for me now. As i feel like i am getting gaslit by the openai api status page on a daily basis

also… to anyone who is saying it… “switch back to completions” is not a f***ing serious suggestion!!! obviously it can be done, but they are of course NOT interchangeable if you are using threads for more than a single run… would require that one implements their own thread and run system for an actual replacement. And even then, you would not have the same dynamic context window management that the actual assistant api threads have.

… but come to think of that… that would be a good project! a model agnostic open-source ai thread and run management client. so just like assistants api, rebuilt, except to implement you would configure the runs to actually go to your llm completion endpoint of choice… (lol just trying to toss this one out into the ether in the hopes that someone will make it for me)

stay hydrated folks!


OpenAI says incident resolved but I am still facing the same issue, not getting any response from the assistants api which earlier used to give response in 2 mins.

1 Like

I am still facing the same issue. Around 20% messages are still failing.


Anyone still facing issue? or is it resolved now. Please confirm.

1 Like

See [here] (Switching from Assistants API to Chat Completion? - #2 by icdev2dev) for a similar idea and ideated implementation.

1 Like

It looks like someone mistakenly changed this topic from API-Bug to API, so I reverted it.

Once we confirm that the issue has been resolved, we will treat it as resolved.

Apologies for the issues here and thanks for your patience – we’ve resolved the incident that caused these timeouts. If you still see them happening frequently, please share a run.id and we’ll take a look.


run_id = ‘run_NNIF0E6I5meslistlmiAWIJN’

Still stucks at ‘in_progress’

1 Like