Has anyone noticed a significant slowness in API queries? I am using the same code and environment but queries that normally takes 1 min now takes 5 minutes. I haven’t changed anything.
Using 3.5-turbo and 16k
#slow API
fallen scaffold
ashen gulch
This shouldn't be the case, can you share some example prompts that you're using?
plucky zinc
This shouldn't be the case, can you share some example prompts that you're using...
Hi 👋
I showed two examples in this chat: #dev-chat message
topaz saddle
I have been experiencing slow responses too 😭
old plume
This is real; GPT-4 is now even faster than 3.5 turbo 16k. The 3.5 turbo 16k prompts that used to take 20 to 30 seconds are now taking up to 80-90 seconds. I haven’t change the prompts so this is issue is not related to the prompts.
zealous gate
This is real; GPT-4 is now even faster than 3.5 turbo 16k. The 3.5 turbo 16k pro...
facts, 3.5-turbo-16k is really slow. I don't know for 4k
plucky zinc
facts, 3.5-turbo-16k is really slow. I don't know for 4k
4k is really slow as well
It's like 10x slower almost, crazy
faint holly
slow for me doring last week too, yup
turbid copper
the longer the context, the longer the wait.
glossy drum
Me too! It's terrible! From 5 seconds to 45 seconds! Using gpt-3.5-turbo 4k.
faint holly
Yup, it really feels weird to see everything green on https://status.openai.com/
the problem is bacl again for me, 60s of total wait (with retries) for pretty short request
@ashen gulch @real musk I'm sorry for tagging you, but I feel that there's an ongoing problem that's not reflected on status.
The most notoriously slow batch of tries of mine hit shortly after 8:00 UTC (~10 mintues ago from time of this post)
using "gpt-3.5-turbo-16k" endpoints, org id is org-nUKjw9MKe38UZr1hI1uX28cN
inland mason
Yup me too. All day yesterday and today too
Requests went from about 10 seconds to over 30 which makes my api gateway timeout
A good way to show this is to ask it for a recipe and it takes about 30-45 seconds to reply
sullen sapphire
Can report the same for both gpt-3.5 and 4 models. We integrated it with streaming, but most of the time, the response is stopped after 4-7 sentences. Its barely usable at this point!
jade violet
I keep getting server is overloaded error
celest schooner
I keep getting server is overloaded error
it is working for me.
inland mason
i noticed that the UI for ChatGPT is significantly faster than the API at the moment. I tried the same prompt in both and the API took about 5x longer
inland mason
any updates to this? I saw the API was having issues but that appears to be resolved. However, I still get very slow responses from the API. I just asked it for a recipe for bannana cream pie and it took 62 seconds to return it.