This is only if you ignore the growing open source models. I'm running Qwen3-30B at home and it works great for most of the use cases I have. I think we're going to find that the optimizations coming from companies out of China are going to continue making local LLMs easier for folks to run.

What hardware do you use ?