I wonder (as an outsider) why would GPT-4o give more accurate responses? The training data is the same, right... so maybe somebody can point this out like for a

I wonder (as an outsider) why would GPT-4o give more accurate responses? The training data is the same, right... so maybe somebody can point this out like for a kid, thank you.