you are viewing a single comment's thread
view the rest of the comments
[–] 3 points 1 month ago

They’re all trained on copyleft data, so they should by all rights be open source. Distilling them into open weight models is the next best thing, I suppose.

  • source