The conventional response to this seems to be "Lower reasoning effort".
I have already tried this.
Like you, I am seeing a 5x slowdown compared to GPT-4O. I have lowered reasoning effort all the way, set verbosity to lowest. Still not getting good results.
The official response seems to be: set reasoning to the lowest and service_tier to priority, and it should work as fast as GPT-4o. But that is not the case.
They are probably still tuning it. Or maybe GPT-5 is not meant to be replacement for GPT-4o / GPT-4.1. Or this is going to be a new normal (slowest).
I thought this was a Cursor problem. GPT-5 is extremely slow at output.
It seems to speed up after a “cold start” though, which seems highly unlikely technically, but that’s what it feels like.
The conventional response to this seems to be "Lower reasoning effort".
I have already tried this.
Like you, I am seeing a 5x slowdown compared to GPT-4O. I have lowered reasoning effort all the way, set verbosity to lowest. Still not getting good results.
https://bsky.app/profile/rickcarlino.bsky.social/post/3lvycs...
OK - so we are not crazy.
The official response seems to be: set reasoning to the lowest and service_tier to priority, and it should work as fast as GPT-4o. But that is not the case.
They are probably still tuning it. Or maybe GPT-5 is not meant to be replacement for GPT-4o / GPT-4.1. Or this is going to be a new normal (slowest).