I am on tier 4 but keep getting an error that I have hit my rate of 300 request/min.
I am using openai.beta.threads.runs.createAndPoll is there a limit on the polling?
I am also running openai.beta.threads.messages.create in a loop. Is there a limit on that?
error: {
message: “You’ve exceeded the 300 request/min rate limit, please slow down and try again.”,
type: ‘invalid_request_error’,
param: null,
code: ‘rate_limit_exceeded’
},
Hi, when sending messages to OpenAI endpoints (especially async or in parallel) make sure there is a min delay of 0.21 sec between requests (must be less than 300 requests per minute).
Thanks @sergeliatko
It will be hard to for my application to ensure this delay without building a fair amount of infrastructure. We are looping through slides in eLearning modules at the user’s request. There could be many users requesting an AI review at the same time and each of them might have 100’s of slides in their module.
We would need to build a job queue, etc to manage this.
A rate of 300/min makes this API almost unusable…
The thoughtful administration of these API call limits, at values that have been updated by OpenAI, to assistants endpoints far below what models accessed by chat completions will provide seems quite intentional. That is perhaps to keep you from deploying it widely while it is a beta product.
The section in Assistants documentation about call limits (originally starting at 60), along with other limitations and plans, has been removed.
1 Like
Well, in my humble opinion, that part definitely sounds like under optimal approach to your problem.
How many steps that require fine-tuned models only do you have in a typical/standard request cycle? If under 3, then by looking through your workflow design you can definitely come up with something that will handle several thousands of regular frequency users with 300 requests/minute…
I would start looking into your workflow first to see why 300 r/m is not enough (what’s the price of your service, is it competitive?)
Olá. Uma dúvida: mesmo pra quem é assinante do GPT-4o existe um limite de uso diário? se sim, qual é o limite? acabei de receber uma mensagem dizendo que atingi meu limite…
Traduzido por ChatGPT:
A resposta é que, ao usar o ChatGPT (que não é a API da OpenAI que fornece códigos de status de erro), o GPT-4 é limitado após 40 entradas por período de três horas, e o GPT-4o é limitado após 80 entradas por período de três horas.
Where is this recorded in the documentation? It’s a bit surprising.
It’s not it’s just that if you have a rate limit and you don’t implement delay between your requests your requests per second maybe too much I’ve ran in this in the beginning but now I’m not sure if this thing is no longer valid or it is me who passed in level 5 don’t have that limit anymore.