This is a follow up to Gio's post on using local LLM models. I show the cost/latency with calling the same model from OpenRouter, and have a blog discussion at Using local models vs API.