Rate limit exceeded API error after topping up balance, disappears after using Playground, then appears after a while

API Bugs
, ,
1

This morning we’ve hit the rate limit and went -0.34$, we instantly topped it up with 100$.
4 hours after I still keep getting rate limit exceeded

last_error: {
      code: 'rate_limit_exceeded',
      message: 'You exceeded your current quota, please check your plan and billing details. For more information on this error, read the docs: https://platform.openai.com/docs/guides/error-codes/api-errors.'
    }

The weirdest thing is, if I go to the playground and play with one of the assistants for 5-10 messages, the API becomes operational, but then again starts showing the same error after 10 minutes or so.
The balance is being updated accordingly, about 30c gone, standing at 99.60$.
I’m guessing it’s a bug on OpenAI side, especially after finding this “workaround” with the playground?

Usage tier 5 by the way.

1 Like
2

Rate limit exceeded. Im guessing your capping per minute.

https://platform.openai.com/docs/guides/rate-limits

3

Thanks for your reply, but unfortunately, we don’t make that many requests, especially to hit the limit for Tier 5.

4

could be tokens per minute that you are capping? are you managing a thread, perhaps the tread history is getting long. when using assistant threads everything you feed in each message is resent with next message and so on which is how thread memory works. or you could be right maybe a billing / account issue.

these are all I could come up with haha. 5-10 message at 30c on which model? I did a whole day on mini and it costed me 11 cents. so 5-10 messages at .30c sounds really high like a lot of tokens.

5

could be tokens per minute that you are capping?

No, it’s still far away from hitting the limit for Tier 5.

are you managing a thread, perhaps the tread history is getting long.

We start a new thread per each refresh of the page, first message instantly fails with that error.

or you could be right maybe a billing / account issue.

I’m guessing it’s this because as soon as I sent a few messages on the playground, it magically worked, and that happened few times, so it’s either massive coincidence each time, or something is wrong

these are all I could come up with haha. 5-10 message at 30c on which model? I did a whole day on mini and it costed me 11 cents. so 5-10 messages at .30c sounds really high like a lot of tokens.

Oh no, I was talking about playground, I sent 5-10 messages to the assistant INSIDE the playground.
The 30c missing is I’m guessing negative balance being evened out, and the rest is gpt-4o-mini usage which I’m using, so the numbers do make sense.

Thanks for your suggestions though!

1 Like
6

Would suggest this as a solution as we had an ongoing loop that was undetected on the usage page until we separated projects and api keys so thanks :smile:

1 Like