Hi, I’m experiencing 429 errors for GPT calls on a long standing (3 months) Tier 1 Pay-As-You-Go account where I’m not hitting my limits. The response given in the body is
{
error: {
message: 'You exceeded your current quota, please check your plan and billing details.',
type: 'insufficient_quota',
param: null,
code: 'insufficient_quota'
}
}
Yet, I have no hard limit set for the organization; I’m not hitting GPT 3.5T or GPT 4’s RPM, RPD, TPM, or token limits; and I’m below my month usage ceiling by more than 50%. Any ideas of what could be causing this?
As part of my testing, I bought $5 worth of credits and got a single call through to GPT4 without a 429, but, after that single call, I started receiving 429s again (on both GPT3.5T and GPT4). I have also confirmed my credit card is valid, my most recent outstanding bill was paid, and replaced my secret key with a newly generated one.
Check the selected organization, by clicking lower-left name icon in the account site. See that you likely have a new “personal” organization that if selected will break everything. See that in API keys you can select different default organizations for all keys, including the one without payment method.
I do see that I have a new personal organization; however, I have the default organization for the API keys to be the “company” one (for my case: spraling). The company organization has a valid credit card in it. However, I am still getting 429 errors for my API keys.
I’ve included an image of the default organization:
You can see if there is someone else helping you spend the money. Go to the “teams” organization, and see if there are members other than just yourself, such as a password cracker that is in your account and spending away.
Then check your usage, see how long it took to spend $5 the day you added it (which now takes a day+ to be seen instead of every 5 minutes)
You might find it wise to log in again with a new private window, but press ‘forgot password’ to get a reset email. Then pick one that’s an original, and that’s after your email can’t be read by such a hacker either.
Then make a new API key and kill all the others.
Would deleting all the API keys break your client app? Then there’s your key leak.
As far as I can determine, no one has hacked my account. Please let me explain some further steps I took to test the problem:
I confirmed that the two other people within the (spraling) organization were both still able to use their API keys. Neither one of them had any 429 errors. Both also confirmed that they had not made any significant calls that would cause rate limiting for the organization (1 call per 6 hours that were under 10,000 context tokens and 1000 generated tokens). I checked this both yesterday and today.
I confirmed that my usage and billing has not changed. I did see changes that reflected other users in the organization that were added into the calculations, making me believe that the usages page is correct. If someone had hacked my account, I believe I should be seeing charges from one of the members increasing to reflect constant 429 errors. (Images included at the bottom of the post)
I deleted my old API keys and started using new ones. (Not that I was expecting anything, but) No changes
I changed the default organization for my API keys to personal, then back to spraling, to reset the established organization and confirm that it wasn’t some issue with the business logic not aligning with the display of the website. (NOTE: this did allow me to submit a single response to GPT4, but nothing after that first request)
I changed the hard limit for the billing from None to the max amount ($120) to confirm that this wasn’t an issue with the system where None was being recorded as a max of $0.00 and preventing any calls from going out.
Weird things to note since yesterday, I have received a few emails from OpenAI today that my $0.00 hard limit was reached, yet I have not made changes since yesterday to change the hard limit. Also there is not record of any limits being reached in the usages page.
I also have tried to reach out to OpenAI’s support, but have received boilerplate responses about increasing my usage limits. Not sure if you have any recs to expedite that process or keep them from sending a basic response then immediately closing the tickets, but any help would be appreciated.
Hey! Never confirmed what the actual solution was or where the problem originated, but the problem occurred after I changed the hard limit on my account to nothing. After I set the hard limit to my max amount allocated and waited a day, I stopped seeing problems.
The page shows that your $10, paid over a year ago, has expired.
A monthly limit of $120 also shows that you have not been updated to the “tier” system for maximum spending.
OpenAI made credits that were purchased expire within a year. They keep the unused money without providing further services.
We will start to see more of this. New accounts created after August 2023 were forced to make prepayments instead of having monthly billing, and then everyone was switched over to this requirement to prepay, purchasing expiring credits, through March-April 2024.
