What's your go-to model in Cursor? A frontend dev's take on Gemini 2.5 Pro vs. Claude 4 vs. GPT-5

Discussions
21

I like gpt-5 more than claude, claude generates too much rubbish code, golang. but the gpt-5 is so slow and always let me confirm….
I learn about some info. I would prefer to use gpt-low, quickly.

Now watiting for gpt-codex.

So much disappointed with claude, so much, so much

22

Really no go-to model tbh. Depends on task at hand

  • simple project, little things - Gemini flash / GPT4
  • structured output - GPT4.1
  • analysis - Gemini Pro or o3
  • code review and brainstorming - Gemini Pro
  • code implementation / debugging - Sonnet 3.7 or 4 depending on how complex the code is. For big refactors 4.1 Opus
  • generating documentation - GPT
  • generating tests - GPT + Claude
  • quick trivial tasks not directly related to coding - Gemini Flash - GPT is annoying with it’s output

I actually use multiple agents for different purposes with case by case selected model.

23

Isn’t that too overwhelming for you when it comes to selecting a proper model for a certain case?

TBH - I personally got “stressed” after just reading your bullet points :joy: No to mention, implementing such a work framework on my end.

BTW. What is your profession?

24

Grok not found…

25

I will always try the models that are released as free for a limited time but only on non-critical projects.

I always end up back with Claude 4 sonnet. The way it tells you what’s going on and the code quality is the best.

26

recently，I prefer to use grok-code-fast for golang, and gpt-5-low for asking question. and rarely use claude.

27

I’m Senior Mobile Engineer (iOS mainly but also CMP / Flutter / Android), but atm I am leading AI adoption and research at my company. I also like to expand my domain knowledge and take part in cross-tech / cross-team initiatives. Tbh selecting model is like choosing framework to go with in frontend dev language. Adaś, po prostu nie bój się próbować i zmieniać modele, to tylko narzędzia ;p