Local LLMs are much better than their reputation

I always read “well, if I only could run the 600b model, it would be usable, 70b and 8b are just bad.

This is not my experience at all. Especially models that “think” before they write are astonishingly accurate even down to 14b.

I always think “damn, you noticed THAT, wow”.

Comments URL: https://news.ycombinator.com/item?id=43058718

Points: 1

# Comments: 0