Is there still a point to fine-tuning local LLMs?

I've been getting more into running LLMs locally, and I'm starting to wonder if fine-tuning is even worth the effort anymore. Models are getting really good, context windows are huge, and with things like RAG,...