Hi everyone,
I usually use Cursor as my main tool and haven’t tried Claude Code before.
Most of the time I work with Claude Opus or Sonnet models.
What has been your experience with Claude Code?
Is it true that if you mainly use Claude models, using Claude Code is more cost-efficient compared to using them through Cursor?
Also, in terms of output quality, do you notice any difference?
Does Cursor produce better results with Claude models, or is Claude Code actually better optimized for them?
For context, I’m currently on the $200/month plan, but it usually gets consumed pretty quickly.
Would love to hear your thoughts and experiences.
From my experience - the Claude code harness is best suited for Claude models. However when I used Cursor after claude code, surprisingly claude models worked too well here too - so it looks like Claude models adapt very well to almost any harness out there (I used CC during sonnet opus v4 days).
Claude code lets you use models from other providers as well if you overwrite a few env variables - so you are not vendor locked to Anthropic - so thats a good thing recently.
So you are basically paying for the model harnesses - the models can be swapped out by choice. It’s upto you which one you like based on personal and professional preferences.
For me - I like Cursor as it gives me a whole IDE, but I use Cursor only for my work - Personally I use Vscode with Cline (generous free models to use). I’ve noticed Cursor is become more money hungry over the year - they can improve if they focused on creating quality over quantity of work in their product.
I forgot to answer your questions:
Is it true that if you mainly use Claude models, using Claude Code is more cost-efficient compared to using them through Cursor?
Yes, anthropic’s subscription is better if you’re using claude code and you don’t run out of credits as easily compared to Cursor. Keep switching models based on task - Opus or sonnet thinking for planning and haiku for writing code - then you will be good. If you use one model - eg. opus for everything - expect to run out of credits soon. Sonnet thinking is a good alternative for planning over opus - if the plan is not good, only then use opus.
Also, in terms of output quality, do you notice any difference?
Does Cursor produce better results with Claude models, or is Claude Code actually better optimized for them?
Both are same in terms of result generation - if you tell them to refactor code - both will do it with utmost precision because the harness is well adapted for claude models in both softwares.
For context, I’m currently on the $200/month plan, but it usually gets consumed pretty quickly.
Since you are paying $200 for Cursor per month, try using a $20 claude pro plan on the side for the same work. You will be able to get a rough idea of how it works, is it better for you or worse.