I have been using Kimi K2 on their official site and it’s incredible.
GPT, Qwen 3, sometimes even Claude sounds dumb in front of Kimi K2
But please fix the integration with Cursor, I am noticing gibberish output format and some issues
I have been using Kimi K2 on their official site and it’s incredible.
GPT, Qwen 3, sometimes even Claude sounds dumb in front of Kimi K2
But please fix the integration with Cursor, I am noticing gibberish output format and some issues
Hey, thanks for the report. Yes, we are already aware of this issue. I’ve also logged it in our bug tracking system, and I believe the team will fix it.
Hi any updated? I get output with a few words or characters per line, making it very hard to read and use.
Update: I just tried this and it worked:
Is there any way you can format your output in to complete sentences? I see your text as only a word or two on each line? Are you adding returns?
…it apologized and immediately started generating nice output! Hopefully it sticks…
It does not stick. Needs reminding. If there is some sort of integration point where it can be reminded to format output for humans, that would be great.
Sadly Kimi k2 is still not fixed
vertical response still and It just keeps showing “we’re having trouble connecting to the model provier”
c5db317d-fafa-4ba0-bb13-c27f1f91fea3
Any update on this. Is it planned to be worked on? I think this thread was marked solved at the point of we “believe the team will fix it”
This model has been a really great one for when you just want solid tool calls and nothing fancy. After Claude, it does the best job at generating git commits with bulleted messages that are as good as what I would write if I cared to take the time to elaborate in my commit messages
…the other models, especially GPT-5 (which does a 1-liner with \n no matter what I ask it to do), can’t seem to pull this off.