Same. For 'sovereignty ' reasons I eventually will move to local processing, but for now in development/prototyping the gap with hosted LLM's seems too wide.