My organization has been a heavy user of OpenAI’s APIs for a couple of years now. We have a monthly spending limit of $80,000 and have a quota of 1M TPM for text-davinci-003. For months now I have constantly reached out to OpenAI about receiving a rate limit increase for gpt-turbo-3.5, and have not gotten a single human response. This has been very troublesome for us because this means that we are not able to move our services from the Completions APIs to the new Chat APIs.
Does anyone have advice on how I can get my rate limit increases approved or get in touch with someone from OpenAI? Dozens of emails and rate limit increases forms later, we still have not heard a single word in over 3 months.
I think part of the issue may be available capacity, with GPU’s and those qualified to install them and locations with sufficient power and cooling infrastructure to host them being at a premium, it’s a balancing act keeping all of the millions of API users connected and receiving a reasonable service.
Minor rate increases have little overall effect by themselves, but either many such small increases or several larger ones could, might be worth booking a ticket to the DevDay in SF that is coming up and getting some face time.
Unfortunately OpenAI’s popularity has lead to a shortage of humans on the ground to cover the ever expanding number of users, Staff members do check the forum parodically and it is possible that one may pick up on your request but it should be relied upon.
See what you have for gpt-3.5-turbo-instruct. It also seems to have TPM classed the same as other completions. And would knock a huge chunk off of your bill. And you’ve got about three months to do it.
Thanks for your response! They actually finally got back to me yesterday and increased us to the max quota because of our spending tier. I totally understand that the volume is high, was just hoping for some help. Luckily, I think posting here did help.
I don’t agree with that logic. Are you suggesting that early adopters should be forced to spend more money to work with lower accuracy models? If anything, I think it should be the opposite and early adopters should have the smoothest transition to new releases and models.
I’m not suggesting that anyone deserves special or priority treatment. But suggesting that because someone has access to text-davinci-003 that they shouldn’t have usable quota on the newer, cheaper, and faster models does not make sense to me.
I am experiencing the same problem. I developed Scraips, an app which creates recipes images with Dalle 3, but I am hitting the limit very fast with only 7 images per minute. Didn’t think about that in the beginning and now I have to rework the complete app
I am facing the same problem regarding Rate Limits increase.
I should be at Tier 3 since I absolutely meet both of the following criterias, but I am still at Tier 1. I have escalated the issue via chatbot from the help page and am waiting for a response from OpenAI but have not heard back.
$50 paid and 7+ days since first successful payment
$100 paid and 7+ days since first successful payment
If anyone knows how they were able to resolve the issue, please let me know.
Request an exception is a form only available to higher payment tiers.
The primary method to increase tier and model limit is to follow the instructions of how much to prepay.
Changing tiers after a payment finally ages enough takes ANOTHER payment for recalculation. If you paid in that $101, and waited a month, it seems nothing’s going to happen until you place at least another minimum $5 into your account.
This topic originates from before tiers and automated increases.