LM Studio is about as simple as it gets, you do need some GPU grunt. I've got a 4090 so VRAM isn't much of a problem, but you can use quantized models which are like using FP16 vs FP8 which lets you squeeze larger models into systems with less VRAM.
LM Studio can also offload to the system memory if you don't have a lot of VRAM.
It works on Windows and Linux just fine, I haven't tried it on MacOS, because why use my m2 macbook for such a task when I have far beefier hardware in my everyday PC.
I've worked with this stuff in a professional context, I find that Gemini is the best for code coherence. I only know Python (and a limited amount at that) - and it is pretty good at converting Python to Java, but not so good at using libraries like React, Django or etc, probably because I haven't asked it to!
RE: ChadGPT on ChatGTP and the Power of Mind Share