Yes and no. Intelligence scaling with compute makes sense so I doubt the advantage of closed models on large compute clusters will ever truly go away.

But that doesn’t mean smaller models aren’t useful.