Batch api stuck on `Validating` status?

Starting this morning at around 6am (GMT+0800) time all requests to the batch api seemed to be stuck in Validating status, although I was running exactly the same code from previous day.

Checked openai status page and there’s no report of accident, wondering is this a known issue?


More details:

[UPDATE] As of 10am GMT+0800 batch api seemed to work and everything worked fine again.

1 Like

Same problem here. Same code and everything, suddenly my requests are stuck on validating. Up until yesterday everything was working for me

1 Like

Having the same problem too, everything was working a few days ago but now any batch request gets stuck

1 Like

Does open ai support read these topics? Or should we contact support?

We are currently facing the same issue with the Batch API being stuck on “In progress.” Is anyone else experiencing this problem?

Same here; I believe it is simply overloaded by too many batches; yesterday, it completed 299/300 in few minutes for me, but then got stuck for many hours.
Very possible, “traffic jam”, need to wait :wink:

BTW, developers shielded this kind of questions from ChatGPT, it won’t disclose company secrets!

Are there any updates? My requests to the Batch API are still stuck in progress as of today (06/06). Some of these requests only took a couple of minutes to complete three days ago

Same here for me unfortunately, batch stuck at validating for over 12 hours now.


Same here. Even very small batches with a few requests (3) get stuck in Validating status.

Same here. I’ve have several batches in Cancelling and Validating state for ~12 hours now.

Same here. I have 3 batches in validating state since 4 hours otherwise it used to take a minute to validate them.

Update: My batches have finished cancelling/validating except for one that’s currently still in the Cancelling state. This might indicate that even requests to cancel are subject to the batch queue, i.e., all batch requests are in the same queue and can take many hours to complete.

A suggestion to the OpenAI team would be to handle cancellation requests at a higher priority than other requests since it’s fairly confusing otherwise (typically, one would expect cancellation to take place without a huge amount of lag).

Another suggestion would be to show some sort of “Queued” state rather than “Validating” since showing the latter state for many hours makes it look like something isn’t working vs. the reality being that your batch is just queued and not being handled yet (by design).


Same here, validating for more than 9 hours, but successfully completed a while ago. It just took a few minutes for the actual completion, though. The system restored service at around 12pm est for both of my batch jobs submitted 9 and 1 hours ago, respectively.