I can run Mistral 7B, in 3-bit quantization, on my phone, and I get helpful answers for (just today) seeing environment variables in Python.

I have a linux box which i run it locally for batch processing, otherwise i just run it through poe (serves as a good comparative website endpoint for differently models i find.)

Check out LMStudio if you want to try. Open source and quite easy to use, doesn't need a beefy machine.

chat.lmsys.org has a zillion models you can use.