It is not necessary to mix down before exporting. However, if you have tracks that overlap other tracks, it can be useful to mix down before export so that you can normalize the mix to a suitable level (generally that would be just a little below 0 dB).
This is the way I handle it when my project is complete and ready for mix down / export:
Ctrl + A (select all), then “Mix and Render to new track”. This mixes the project down to a single track
Solo the new track, and Normalize to -1 dB
Listen to the track and check that I’m happy with it. When possible, I like to have a break of at least a couple of hours before this step so that I can listen to it fresh. Ideally, this will be the day after I completed the project.
Export the soloed mix track. If all other tracks are muted, then just the normal “Export” may be used (muted tracks are excluded from the export), though I will often use “Export Selected Audio” as a reminder to self that I am exporting only the one track. This is my backup copy.
If I want my final “product” to be MP3, I normalize again a bit lower, say around -2 to -3 dB. This is because MP3 encoding can create peaks that are a little higher than the original. If my final format will be WAV or FLAC, skip this step.
Export the final mix
Unmute and delete the mix-down track (no longer needed)
Save the project.
Note that I am saving the multi-track version of the project. If in a couple of days time I realise that the mix is not quite right, then having the multi-track version means that I can do something about it.