Has anyone here been able to use small (~7B params) open source models for anything remotely useful? I mean it’s incredibly cool that you can run them locally on your phone but they’re just dumb as rocks. The only example I can think of is the model underlying autocorrect in iOS which AFAIK runs on device.

Sure that example was not great. Here is example of the most basic mathematical problem solving. It almost gets it but then fumbles the ball in the end. I think these small models are still at the “fun to play around with” phase (which is totally fine!) but not yet at the “useful tool” phase where GPT4 is.

Computational biologist, data scientist, digital artist | 
https://clauswilke.art/links | Hosting /ai-art & /clauswilke

I mean, I don't know what TAWNY is either. So I'm dumb as rocks also? (Maybe I am. Maybe I would say google it. But out of the box, I'm no smarter than this model apparently.)