Comment by julianlam

Base qwen 3.8 27B was so bad at over thinking that I couldn't work with it day to day. Swift solves that.

Perhaps I'll give bottlecap AI's fine tune a go.