Did anyone apply for Azure OpenAI recently? Can you share your experiences?
I’ve looked at their application form. They say they provide Azure OpenAI for enterprise customers only. They require a business phone number.
I have a business license in my country, but I’m only a guy at home. Do they call you at the provided phone number to check on you?
If any of you are a business with office, you should apply for Azure OpenAI asap. This is not a service for business. This feels more like a field of experimentation.
For me, the slow down is also occurring but with all models. In the playground and via normal API access. Last time I used it (about two weeks ago) everything was at normal speed. Now, for about 250 total tokens it’s like 6-9 seconds, before 0-3 seconds.
I am also having the slowdown issue. With gpt-3.5-turbo, responses are taking 3-4 minutes (~3500 total tokens). My larger requests against gpt-3.5-turbo-16k-0603 just time out. This issue is new since the end of last week.
Have tried this both running locally, as well as from AWS EC2, same results.
It looks like the symptom is spreading to more people. Could you guys report this symptom to help.openai.com? Use the chat button on the lower right corner. Maybe link this post along with it.
They give a response, even if it’s only a standard answer.
I already did it. Maybe more people reporting the same issue would help escalate it.
I found a pattern. My account, which was made on July 1st, has normal speed as of now. Accounts created after that day are all slow.
It looks clear to me now. OpenAI is directing recently created accounts to crowded nodes. That’s why the oldbies in this forum are seemingly oblivious to this issue.
This is different then and a completely different problem, this has to do with the date the account was created on. That is so odd. Please supply the version you were using in playground. Because this is different than the current topic please create a new topic and list your issue.
I’m also experiencing significantly longer API wait times for the past 8-9 Days.
A request that would normally take 15-30 Seconds is now taking upward of 75!
Hope open AI can communicate better on this rather than leaving us in the dark, especially because these longer wait times are a significant UX barrier in already launched apps.
I am having the same issues. We are seeing around 38-45 seconds or timeouts on things that used to take around 10-15 seconds max. It is making it unusable for our application.
It looks the speed depends on how much you spend.
As we are in implementation and testing stage, spending nor more than $10 a month, it is very slow.
A friend of mine, who spend around $500 a month on prod load says it is super fast.
It looks OpenAI have different infrastructure for different accounts based on their stage (pre-prod or prod), and this is determined per spending. Anybody can confirm?
The observed difference in response times could be attributed to various factors such as network latency, server load, or other technical considerations. To investigate and address this issue, it would be advisable to reach out to the support team or contact OpenAI directly with information about the specific account and usage details. They will be better equipped to assist and provide insights regarding the slow response times experienced with the GPT 3.5 Turbo API.
So, randomly everyone is noticing a 3-5-fold increase all of a sudden and it is something to do with our individual accounts with everything else at parity? I feel like there is something more to this.
This is certainly not being experienced by everyone, this is a plot of the time taken to generate 256 tokens by GPT-3.5 and GPT-4 over the past week, there are some blips from infrastructure issues and the normal issues of internet connected API’s but in general the performance looks to be unchanged.
My account was created well before July 1st and is experiencing this slowness. I also created a new organization and am also experiencing the same issue with the new org and keys. Based out of Toronto, with great internet speed.
Of the people experiencing the slowness, what is your monthly usage limit? Mine is $120 and I’ve requested it raised as I will be boarding many new clients soon on a service which depends on this API. Not sure if there’s a correlation to that spend limit or possibly location.
