You’re welcome — it’s too late for me to edit my wording but hopefully you understood what I meant by “1/3 to 1/4 slower.” That is ambiguous, instead I should

You’re welcome — it’s too late for me to edit my wording but hopefully you understood what I meant by “1/3 to 1/4 slower.” That is ambiguous, instead I should have said that models that fit in memory on both take 3-4 times as long on the M2 as they do on the 4090.