32 and 64 bit floating point mathematics are used for ternal processing of audio and or fx processing.
While 32 bit files can be exported they are no longer floating point.
Also while 32 bit files can be exported the only use for it would be to import into another program for further manipulation. Why? As yet there are no devices capable of playing back 32 bit files. There is no need. From an audio standpoint while internal numerical calculations can be more refined with less rounding errors, by the higher bit rates, for the purpose of listening there is no point for that resolution.
You'll get better results by increasing sample rate and that is a hot bed of subjective thought that can lead to months worth of arguments.