What model are you using nowadays?

Discussion
1

Curious what models people are using nowadays.

I’ve mostly stuck with Claude 3.5. I feel like 3.7 makes too many extra changes, overcomplicates an otherwise simple solution, does weird terminal commands, and it feels like it doesn’t read as much context as 3.5 sometimes.

Perhaps I need to adjust my Cursor rules specifically to adjust 3.7 to behave more like 3.5.

I’ve also heard some people prefer Gemini 2.5.

So what models is everyone primarily using, and do you use any rules to adjust a models behavior?

1 Like
2

I just started using gemini-2.5-pro-exp and so far it’s amazing. Feels like a major upgrade from Claude 3.7. I agree that Claude 3.7 felt too excitable, often getting carried away. 3.5 was reliable but just not very impressive. By contrast, Gemini reasons things out and nails the assignment.

I also want to try out o3-mini but haven’t gotten to it yet. This benchmark is handy: LiveBench

Claude used to top the charts for coding, now they’re way down the list.