Agree on open source, disagree on token efficiency. OpenAI's model are by far the most efficient. Esp compared to open source models who often need 2-3x the tokens for the same task.

→ View original post on X — @steipete,