I Use Both: Claude and ChatGPT

People sometimes ask me, "What do I use? Claude, ChatGPT, or maybe Cursor, maybe something else?" So the answer is the following. It all depends on the budget. I initially started using just ChatGPT. I was paying about 20 euros per month. Then I understood that the amount of tokens which I can consume or the amount of inference which I can consume with a Codex solution from ChatGPT is quite limited. It was at least not enough for me. I was comfortable to maybe not switch to the higher plan which ChatGPT is offering for more than 100 or more than 100 euros per month.

It was not really comfortable for me to pay this amount of money for my personal projects. And as a result, I opted into Claude Code, into another solution which costs approximately the same, I think like 17 or 20 euros per month, which gives me actually a bit more amount of inference than ChatGPT. So what I'm currently doing, I'm doing the following. So when I'm working on my projects and just for the context, I think on the personal projects I'm working, let's say one, I have an inference of one or two hours per day.

So ChatGPT or Claude is doing my projects and consuming their resources on my tasks. So for this type of, for this amount of work, two solutions, one from Claude and one from ChatGPT is actually enough. I'm using normally the latest models like Astra 6 now or maybe I'm switching to the models before, the latest models if I know that the task is not that complex. And what is very important, if you are working on the same project with these two models, so with these two providers, you should use GitHub version control so that when you finish with one model, you just basically have everything committed and maybe even pushed to GitHub.

And after that, you are switching to another model and continuing working with this other model on the same repository, maybe in another branch. And it works, it works personally for me. If you can actually switch like or separate the projects in one one project you are working just with Claude or another one with ChatGPT, it is even better. The only thing I will recommend you to be aware of on all these plans, for example, for ChatGPT, actually on all the plans, if you're using the personal version and not enterprise version, by default, your data can be used by OpenAI also for training.

So make sure that you go to settings and you switch off this default setting so that this data is not used by OpenAI for training of other models. So that's it, that's my setup, Claude and ChatGPT. Codex, I use both depending on the project. For more visually important tasks, I use Codex because I found that OpenAI is better for these tasks and for routine things and the things where a lot of compute is required, I use Claude.