Max thinking doing worse than non-thinking is the funniest part of this benchmark, models really do overthink now.

→ View original post on X — @whats_ai