openai.error.InvalidRequestError: Too many parallel completions requested

I’m getting this error when I try to do an openai.Completion.create:

openai.error.InvalidRequestError: Too many parallel completions requested. You submitted 24 prompts, but you can currently request up to at most a total of 20). Please contact us through our help center at for further questions.

The code was working initially, and as I tested the core more, I eventually received this error. I waited 12 hours to see if the issue resolved itself and I’m still getting the error. I’ve googled around, and I can’t find this error anywhere else. Does anyone know what this error is and how to fix it?

Error message above seems clear. If it was working before, and not working now, the logical conclusion is that OpenAI made a change and limited the parallel completions to 20. After all, it’s a BETA and engineers are going the make changes, for sure.

The way to “fix it” it to change your code to submit 20 prompt in parallel, not 24, right?

My code is submitting one prompt at a time. It seems like there’s some kind of “close” completion that’s necessary for the API, but I do not see any equivalent.

Thanks for telling us that, just now :slight_smile:

So, I have no idea as I have never seen this error. Have submitted a number of completions today and some failed due to “load” but I just resubmitted a few seconds later and all was fine.

Hopefully, someone else has seen that error message before and can help you @eljefe6a

Maybe it a “copy-and-paste” error message as the OpenAI team works hard to update all the error messages under the gun, as they are now, and this error message is wrong (copy-and-paste error)?

I’m still getting this error 4 days later, and OpenAI’s support hasn’t contacted me. Has anyone else hit this?

@eljefe6a This can happen even if you are sending one request but in split mode. e.g the following code will send more than one request as it is splitting it while sending the request

food_dataset =
    lambda x: {"parsed_data": get_spices(x["Recipe"].split())}

however, the following code will send only one request

food_dataset =
    lambda x: {"parsed_data": get_spices(x["Recipe"])}

I double-checked with logging output that only one completion is being submitted at a time.

okay, is it possible for you to share the code?

@eljefe6a Hello! How are you? I’m receiving the same error! and it’s taking long time to resolve. Could you please tell me if you got this resolved? Thanks in advance.

I never did find a solution. I think it’s a bug in GPT with the tokens or the number of tokens. I contacted support, but they weren’t helpful.

maybe some other app is using your api key or another api key in your account, double check usage page

@eljefe6a Okay. Thanks for the info. @Gaouzief Currently, we are only using one model where the split the paragraph and send prompt requests. sometime it says too many parallel completions and default is 20 only. I have created ticket to support.