[Critical] Over 25% Assistant API Request Timeout Randomly

No, I checked OpenAI’s status, and they are stating “All Systems Operational”.


i think they are all still in bed, whilst my reputation lies in tatters :no_mouth:


I am getting time out on more than 50% requests now.

rest in pieces, my dear sweet professional reputation! :rofl:


With gpt-4-vision-preview, I’m also getting about 25% of my requests failing with:

error sending request for url (https://api.openai.com/v1/chat/completions): operation timed out

1 Like

Unfortunately we’re dealing with the same thing. Just lost 2 weeks worth of work since we now have to roll this all back.

The worst part is when the status site says that “all systems are operational”, I don’t even have a way to explain to clients that the system is actually failing, as many here say “my reputation was lost in a second”, and even though is a “beta”, the truth is that we paid a lot for this beta, and if the risk worried them they should not give us 10,000 RPM to use it


Same issue - the API is non responsive at the moment.

I am having the same problem. Almost 90% of my calls are either timed out or canceled.

To overcome this, I have at least 20 retries if something fails and I have a timeout of 5 seconds on the httpclient to minimize the wait in those rare cases the api times out.

Same here… also can’t upload files to the assistant in the playground even! Trying to build a business here!!!

When I console run, this is the point where it stucks:-
id: ‘run_4bG0wGmHZJqqdO0u8bGoemYb’,
object: ‘thread.run’,
created_at: 1709307122,
assistant_id: ‘asst_HFrxlpyT1MbVYzejPzhOfTaP’,
thread_id: ‘thread_SlamGiMcWsyw4ZN2kMyowuBN’,
status: ‘in_progress’,
started_at: 1709307122,
expires_at: 1709307722,
cancelled_at: null,
failed_at: null,
completed_at: null,
required_action: null,
last_error: null,
model: ‘gpt-4-0125-preview’,…}

Then I get “completed” status out of nowhere taking huge amount of tokens which is 3x times the tokens being used earlier to get the response.

You’ll have to give them some time to wake up and flip the “is broken” switch.

1 Like

What a pain, I was at an AI conference yesterday and no wonder everybody is using screen recordings…


yes. if they could at least update the status so that we can pass this on, so we don’t look like we are completely ineffective. as others have said, we are paying for this. and yes it may be beta but its all bloody beta and we are going to use it no matter what.

It’s been like this for 7 hours now.


I gues the problem is the lack of transparency. There is no incident in the open AI dashboard: OpenAI Status - Incident History

Same issue here.

I think by forcing developers to poll the result of a run, many backend/clients are stuck in endless polling, that lead to a flood of requests taking their system down.

Hopefully it forces them to prioritise streaming of the run request, so we can get rid of polling.

Hello everyone,

I’m currently encountering a recent issue with integrating my application with the OpenAI Assistant API. Since this morning, I’ve been facing difficulties where requests to the API simply stop responding. It seems like communication is initially established, but then the response halts, as if the API is freezing. Everything was working fine yesterday, but today I only receive timeouts due to the communication failure with the API.

Here are some excerpts from the logs I’m getting:

2024-03-01 17:43:37,664 - DEBUG - Starting new HTTPS connection (1): api.openai.com:443
2024-03-01 17:43:37,660 - DEBUG - https://api.openai.com:443 "GET /v1/threads/thread_LcGgkHPIGmKZi6PNqSN2Vz1K/runs/run_G9LDQDexJdeq3IoEmbXAv9N3 HTTP/1.1" 200 None
2024-03-01 17:43:37,664 - DEBUG - Starting new HTTPS connection (1): api.openai.com:443

I’ve already checked my credentials and internet connection, but the issue persists. It seems like something has changed on the API side. Is anyone else experiencing this problem today?

Appreciate any help or guidance on resolving this issue.

Thank you!

It’s not working even in their assistant UI, with their automatic apis. So yeah… it’s a bug

1 Like

I was going crazy trying to figure out what else I had broken! But yeah, same issue as everyone else. Using the Python SDK, I’m having timeouts creating new assistants with

return client.beta.assistants.create(name=name, instructions=instructions, tools=tools, model=model)

I’m even having trouble deleting assistants with client.beta.assistants.delete(assistant.id)

The Assistant API is great, I hope it continues to be supported! Also, more granularity around these types of errors would be helpful. The official OpenAI status page generally doesn’t mention API specific issues.