Local LLMs are moving fast. In the six or so months I’ve been running them on my own PC, what’s actually available has changed a lot, and not always in the direction of bigger. The people building these models understand who’s actually running them, which is mostly normal folks with consumer hardware, not labs or server farms.