Addressing GPT-4 Rate Limit Concerns for Our SaaS Serving Hundreds of Organizations

API
, ,
1

Hi all,

We are developing a SaaS that will work with ChatGPT and offer it to our customers.
We plan to offer this service to several hundred organizations with ChatGPT’s GPT-4 model.
Since each organization has hundreds to thousands of employees, it is very likely that the GPT-4 model rate limit of 40k TPM and 200 RPM will be reached frequently.

Normally, we think that users are applying for the GPT-4 model on Waitlist. For a SaaS provider like us, the challenge is that the rate limit is quickly reached with a single API key because the users are in a state of flux.
Is it necessary for us SaaS providers to register on behalf of our contracted customers with Waitlist and issue hundreds of API keys for them to use?

Thanks,
Kahori

1 Like
2

It is up to you how to handle API keys and accounts for customers, it is valid to hand off keys to customers after the project is finished and for you to handle them as an ongoing contract. YOu can apply for an increase to your rate and overall limits if you have supporting documentation to show high usage, you can find that and a full explanation here :OpenAI Platform