McCaleb
January 17, 2026, 7:26pm
1
Cursor IDE
On rare occasion when the model begins writing the plan, it outputs like 1 token every 3 seconds.
Use plan mode.
Windows 10/11
Version: 2.3.38 (user setup)
GPT-5.2 High Fast (though I don’t know what models I’ve used before when this happened).
b45969a8-7e42-472c-902c-3130edeff2cf
No - Cursor works, but with this issue
2 Likes
deanrie
January 18, 2026, 10:00am
3
Hey, thanks for the report.
Plan mode currently saves the file for every character, which causes rendering issues and slows streaming down a lot. Other users have reported something similar, the difference can be anywhere from about 10 seconds in Ask mode to 5 to 10 minutes in Plan mode for the same request: Plan writing is extra slow
Good news, the team is working on a fix that will ship in Cursor 2.4 later this month.
1 Like
Gulzar
January 20, 2026, 3:29am
4
are you seriously not rolling this back? don’t you have a feature flag? this is THE core functionality of cursor and is unusable
McCaleb
January 20, 2026, 1:10pm
5
This is part of the price we pay for being on the cutting edge. They’ll work it out.
system
February 11, 2026, 1:11pm
6
This topic was automatically closed 22 days after the last reply. New replies are no longer allowed.
