Interesting that it wasn't more efficient than 5.3 codex, like OpenAI said it should be

→ View original post on X — @petergostev