There is a familiar shape to this. A capability that used to require a data centre becomes something you can run on a machine in the spare room, and a hobby is born — the same arc that gave us home NAS boxes and self-hosted everything.
Good enough, and yours
The open-weight models you can run locally are not the biggest models in the world, and they do not need to be. For drafting, summarising, coding help, and answering questions about your own documents, a model that runs on a single consumer GPU is well past the threshold of useful.
What you get in exchange for a bit of setup is total ownership: no rate limits, no usage meter, no terms of service that change under you, and nothing leaving the house. For a certain kind of person, that trade is irresistible.
The appeal is not that it is the best model. The appeal is that it is yours.
It is fiddly, it is a little obsessive, and it is exactly the sort of thing that turns into a mainstream default a few years later. Home servers did. This will too.