Any logic on GPT-4 API quota increase?

As I’ve understood for now, there is no way to request a quota increase manually, especially for “personal” accounts (which has been registered without providing a company).

However, OpenAI page says that they are automatically adjusting API limits. Do this correlate with your personal average usage of a model or they just lift the restrictions on special categories of accounts without taking into account specifically my actual usage?

  1. In other words, if I will load up my quota on full 24/7, do this increase my chance of getting increased quota added? Maybe any other ideas about quota increase logic
  2. If not, is there ANY other way to increase my GPT-4 quota? Maybe upgrading an account to some corporate version with actual company registered or sending them my startup idea somewhere?
1 Like

Hi and welcome to the Developer Forum!

There is no way to increase your GPT-4 rate and usage limits other than to regularly demonstrate you are using the API close to your limits.

There are no corporate accounts for sale, simply use the account regularly and make use of the moderation endpoints to ensure you are performing due diligence on both your input data and that generated by the model.

Thank you for your reply. So you can confirm that daily usage affects the probability of automatic limits rising?

That I didn’t quite understood. GPT-4 itself is censored, isn’t it? What endpoints you are referencing to?

Yes and no. For example, when you use ChatGPT and submit something that breaks OpenAI’s rules, you get the typical “This message does not accomplish…”. But, via the API, YOU are the responsible for moderating messages sent to the API.
Moderation docs

3 Likes

GPT-4 will do censoring, but you should send text from your users to the moderation endpoint first, and use the trigger levels it produces, both the true/false flags and you can use the floating point values to build your own moderation levels. Also do this for output generated by the model, this ensures that you catch terms of service breaking input and output and that you are making an effort to comply with the terms of your agreement, things like that show you would make a good candidate for increases.

3 Likes

So, if I understand correctly, first I should send my query to the text-moderation-latest model and then (if it’s good) proceed with submitting the input to gpt-4 model? Big thanks for that highlight, I’ll definitely implement that feature! But about the rate limits itself, anything else I could do to receive an increase?

I was able to request a quota increase through the form linked to on the “Usage limits” settings page (under Billing) and I recieved it a few days later. But I’m not trying to launch a public application.

1 Like

Thank you! Could you please tell what was your quota usage the last month before application (as it says in that Google.Docs form)? I mean how much of your available limit did you use to get the increase.

I used about $110 of my $120 limit, and my request to increase that to $240 was accepted.

1 Like