> Most app features don’t need a model that can write Shakespeare, explain quantum mechanics, and pass the bar exam. They need a model that can do one of these reliably: summarize, classify, extract, rewrite, or normalize.

> And for those tasks, local models can be truly excellent.

100% true and I use them for this. But the open-source models seem to be drying up unfortunately. There never was much incentive for the big players to train a model and give it away for free, it was mostly virtue signalling and advertising for their knowhow. The AI "race" seems to have entered a new phase that's more on clamping down costs and making money and this doesn't fit in well.

I hope good local models will still appear but the days that there was a new groundbreaking model for download every couple of weeks is over :'(