When working with LLM, does it understand you? When I am working on my project https://mathedu.online, I tested several models. Right now I am working with Gemini 2.5 Pro. It is pretty good, but in some cases I needed to be more strict.
On the other hand, working with Claude Sonnet is like when you meet you soulmate. Understands almost everything what I say. I mean do not give up when LLM is not perfect. Try another one.