Comment by jefffoster

Does anyone have any intuition about the how reasoning improves based on the strength of the underlying model?

I’m wondering whether this seemingly underwhelming bump on 4o magnifies when/if reasoning is added.