Loading assistant with 20 rps

Hello! How much load can openai handle on one assistant?

Several times we observed delays of up to 10 minutes with not more 20 rps

The assistant is not some computation entity that you have created.

It is just a collection of instructions, along with threads, in a database.

I cannot see where there would be any bottleneck just for you. You are sharing computational resources with thousands.

OK. When making requests to the API, there is a long response time of up to 10 minutes, I would like to understand the reason.