429 errors - Has OpenAI billing changed?

I have an application that has integration with gpt-3.5-turbo-instruct, but I haven’t used it for a while (a few months?)

Just tried it today, and all queries are failing with a 429 - insufficient quota error.

For the previous year or so my credit card just got billed every few days with whatever I used, but now there seems to be some sort of pre-paid credits system with auto-topup.

Is this new? Where can I find out about it? And how long does it take for credits to be usable after I pay for them?

Any help much appreciated,

David

429 - You exceeded your current quota, please check your plan and billing details Cause: You have run out of credits or hit your maximum monthly spend.
Solution: Buy more credits or learn how to increase your limits.

You can check here for API…

https://platform.openai.com/docs/quickstart

1 Like

Thanks, but I don’t know what you mean. I don’t understand what quota I’ve allegedly exceeded - it’s not my RPM or RPD because I can’t get a single request through, so my RPD and RPM are both 0 :joy:

My balance is positive, and auto-topup is enabled … so no clue what this mess is all about. Looks like OpenAI ran out of money and got the employees’ kids to build an e-commerce front-end for them.

I’m guessing this is an incident. I’m seeing the same 429 on all my requests.

4 Likes

Chatting with gpt-3.5-turbo-instruct seems fine.

(this AI doesn’t know the right answer, but doesn’t refuse producing a plausible one)

I would investigate:

  • Whether you have enough prepaid credits to fund new queries. Credit balance:

https://platform.openai.com/settings/organization/billing/overview

  • Whether you have set a monthly budget limit that has been exceeded:

https://platform.openai.com/settings/organization/limits

  • Whether overall organization monthly limits are hit (expand “view monthly bill for the whole organization”.

https://platform.openai.com/usage

New payments may simply take hours to restore service despite them showing as a credit balance. A continuing “bug”.


This error code has also been reported, with poor message reason, when using an API key without project rights to the endpoint or API key-level blocking of the model. New features. Generating a new API key in a project should free you of any migration hurdles with these backend changes.

1 Like

I think this is the most likely explanation. I can’t find any information about this new “credits” system but everything is reported as positive, and they weren’t slow to take the money off my credit card so that bit’s working :joy:

A year on from when I first integrated OpenAI it’s a cute toy, but it’s sadly nowhere near reliable or resilient enough to be a business-ready service. I suppose it’s a victim of its own success.

There were several rounds of emails sent to the account’s contact in migration of accounts from monthly billing to the new prepay-required system, starting in batches of accounts in March. Some didn’t receive their communications, apparently.

The final email (links with tracking removed):

We’ve updated the billing system for your [OpenAI API account]. Instead of receiving a bill at the end of the month, you will now pre-purchase credits to use the API. This change should enable you to access higher usage tiers more quickly. You can add credits to your account by visiting the [billing page]. To learn more about prepaid billing, please see this [help center article].

To continue using the API, please add credits to your account by visiting the [billing page].

Please note that this change only applies to your OpenAI API account and Playground. It does not affect ChatGPT Plus subscriptions.

Best,
The OpenAI team

3 Likes

Ahh! That makes sense.

Nope, I just checked my spam folders etc., and never received the emails so this is the first I’ve heard of it. Much appreciated.

I guess it’s just a bugged rollout, probably get fixed at some point in the future.

Thanks for the solution!

1 Like