Interesting to see the discussion here, around why would anyone want to do local models, while at the same time in the Ollama turbo thread, people are raging about the move away from a local-only focus.