Comment by int_19h

That's where you spin up a local LLaMA instance. The largest models that are still runnable on consumer grade hardware actually beat GPT-3.5 at this point. And there are numerous finetunes all over the "spiciness" spectrum.