Comment by Ey7NFZ3P0nzAe

Well, behind "models" not "langual models".

Of course models purely made for image stuff will completely wipe it out. The vision language models are useful for their generalist capabilities