I use local models to explore, hosted models to refine. I somewhat envy those who can sustain local models (q8 120b+) running as a hobby.... for me, the practical path is a better SearXNG setup and knowing my routes forward.