Rate limit

At a glance

The community member is using the free credit from OpenAI and is experiencing an issue with the "gpt-3.5-turbo" model, where they are receiving a "Rate limit reached for default-gpt-3.5-turbo" error with a limit of 3 requests per minute. This is unusual as they have used the same account, organization, and key with more than 3 requests per minute before. The community members suggest that the community member may need to add payment information to their account and that the limit of 3 requests per minute is very strict. They also mention that the community member can set spending limits on their account. Other community members provide code examples on how to use the "gpt-3.5-turbo" model instead of the default. One community member also mentions that the "gpt-3.5-turbi-0613" model has a limit of 60 requests per minute, which may work better for testing purposes.

TTesterMan

Hi everyone, I have a question about openAI, I am currently using the free credit since I am still in testing phase and today I am having a problem. I use "gpt-3.5-turbo" for the queries and today it started tellig me "Rate limit reached for default-gpt-3.5-turbo" and it says that the limit is 3 requests / min. I have never had this error before and I always used the same account with same organisation and same key with a lot more than 3 requests/min. Does it comes from a openAI update or something?

4 comments

LLogan M

You'll have to add payment info. 3 requests/min is super strict, not sure how they decide when to hit people with that

You can set spending limits though on your account. I have mine sent to $20/month lol

eethan

How do you use gpt-3.5-turbo instead of the default? I did this but my code seems still running davinci-003:

service_context = ServiceContext.from_defaults(
llm=OpenAI(
model='gpt-3.5-turbo',
),
)

docindex = ListIndex(nodes, service_context=service_context)

LLogan M

Missing a few lines of code

Plain Text

from langchain.chat_models import ChatOpenAI
from llama_index import LLMPredictor

service_context = ServiceContext.from_defaults(llm_predictor=LLMPredictor(llm=ChatOpenAI(model_name="gpt-3.5-turbo", temperature=0)))

TTesterMan

Oh ok so it is a new thing😱
But I have just seen that gpt-3.5-turbi-0613 has a limit of 60 request per minute, that should work for testing. I am currently using that. but if you are a pay-as-you-go user the limit is stil 60 for chat for the first 48 hours and then 3500 RPM or 90000 TPM.

Add a reply

Find answers from the community

Rate limit