Benefit of recording 32-bit audio?

If it was just 32 then yes, but when it’s float? More headroom, less errors?