raven walrus
Is it normal for the GPT3.5-turbo model to be performing slower than the text-davinci-003 model? I have a webapp that is utilizing davinci-003 at the moment, but noticed when I switch to GPT3.5 Turbo, that the response time actually becomes longer, despite it being the "turbo version"
Wanted to leave this here and ask if anyone else has experienced this same issue or whether this is a known thing that GPT3.5-turbo is slower.
distant arch
Yes, the response time is flaky and can be significantly high some times during the day
uneven relic
Is it normal for the GPT3.5-turbo model to be performing slower than the text-da...
I just converted to model GPT 3.5turbo yesterday. And yea it's very slow compared to text-davinci-003. I even tested it " what is 1+1?" it took longer than 10second just to reply "1+1 is 2"
raven walrus
just bumping this ^ i noticed its severely faster now, so I switched my model and noticed that its also far more cost effective too @uneven relic
uneven relic
just bumping this ^ i noticed its severely faster now, so I switched my model an...
The speed is consistent for GPT3.5 turbo model at your test?
raven walrus
its as fast if not fsater than davinci-003 in my experience