I am have been seeing this issue for the past 12 hrs. I am a free account, but was considering moving to a paid account. I’ll be waiting to pay until I can at least see a request go through!
Someone above stated earlier that had success when changing models. Unfortunately, I see the same behavior regardless of using 001, 002, or 003.
Same issue here. Pay as you go, old account… should have 3,000 requests / minute as limit. Yet, keep receiving 429 - too many requests error, after only a couple of dozen requests in a row
Same here. Staff stand by to use the new functions built on top of this and it’s useless. Also a paid user. - I would say its getting worse.
Same here, I have paid plan and it’s getting worse, sometimes it’s smooth and sometimes all my API calls have 429 error as a response.
I am using cache in order to reduce the number of API calls, and I checked the number of call per minute and I am far far away from the limit.
VERY intermittent yes.
Same here, GTP3 being a crucial part of my app, this will hopefully get fixed soon 
Same here - clearly an issue. Please fix asap
Please fix this, I’m tired of sending emails to our users saying that our servers are overloaded when they’re not…
Same here, really annoying. Haveing a paid account and maybe sending 1 or 2 requests per minute for testing 
Sometimes its running, sometimes it fails
You can add a test condition to your code and when the API is overloaded and fails, you can wait a few “random” seconds and resubmit.
It’s just a few lines of code. Of course, you will need to parse the API error messages and make sure it is a type of “overloading” issue, testing it, etc.
Maybe on every retry send an email to OpenAI support. Complaining haha 
It feels a bit wrong to implement a retry-mechanism when the API returns “too many requests” 
but I guess it will have to do for now.
Make sure your account is attached to a card even if youre under the $18
I implemented retry logic in my code with a random time out and varied the davinci model, but still getting the too many requests. are you a paid account?
Similar issue here. Any requests through the API gives the same 429 error. Even sending a single request with the API playground gives the ‘You’ve reached your usage limit.’ error. I have a paid account, and am not even close to the usage limit. I tried creating a second account with a new email, credit card and phone number and got the same error.
Anyone else completely locked out of the API?
Just getting set up so could be a noob issue but my initial request are returning status code 429.
If you search the forum (the big magnifying glass icon on the top right of every page), you will see that this topic has been discussed to death already:
That tends to happen when an issue keeps getting
worse instead of getting resolved
You’re not alone. Nowhere near usage limits, and OpenAI is charging us for PROMPT tokens even when completions fail. Ridiculous.
Ok. It looks as though we are never going to get a response from support. I’m sure they are overwhelmed, but this issue is very pervasive. So I have some questions if anyone would be willing to answer:
Perhaps if each user lists their specific issues, when a support resource is able to look at the stats they will be able to more easily isolate the issue.
My answers:
5 weeks, give or take.
Yes, my completion success rate is over 99%. The only problem I have ever had was when I called the API in a loop without any delay. Now I always add delay in my loops which call the API.
The API works fine for me, especially considering the API is considered a “research beta”, to be honest, and the API has never stopped working but it has failed a handful of time due to system loading issues (expected during such media hype and exponential growth period in a beta release). Very sorry to hear to you are having this problem, and I hope things improve for everyone soon.
I have sent many requests today with 100% completion rate. Yesterday, I sent many fine-tunings, completions and embedding request without any noticeable issues, except of course fine-tuning is, understandable, a slow process.
None.
I am using credits generously granted to me by OpenAI (Thanks OpenAI); and am reluctant to move to a fully paid account until the growing pains are better. My experience over the years in tech is that if you are “too early an adopter” you put yourself at risk of being disappointed and frustrated, so I practice “late, early adoption” or even “delayed adoption” depending on the situation. I try to keep my expectations low with beta releases and generally try to participate in helping debug the system, patiently.
N/A
