Chiptune formats for retro videogame music can be very efficient. Just picking two with particularly good music, I have a 21 KB (0.02 MB) file storing 28:30 of music and 4.72 MB of files storing 1:54:48 of music, both at source quality.
The catch is that they are designed exclusively to rip chiptunes from retro videogames as close as the format designers and player coders could manage to the original. So even the oversized ones like the 4.72 MB of files extracted from a 3 MB game are going to be far smaller than a general use format like opus. But you can't encode your own music in the format without going to massive effort to code it like you would an authentic chiptune, and you're unlikely to like the results.
Those are SPC files, and that particular example was one rip of Final Fantasy VI (III)'s soundtrack.
Unfortunately, it only handles music embedded in Super Famicom/Super Nintendo games. To convert your own music to SPC, you'd have to rewrite it for the SNES sound chip.
The even more efficient example was Mega Man 3. The standard rip format for NES music is far more efficient but also far more complex, requiring specialized skills to rip instead of a copy of ZSNES and a fast finger on the F1 button.
Edit: the standard rip format for NES music is NSF, but an expanded version NSFe is better if you can get it because it supports metadata like song names and lengths.
Everything filed under "Chiptune", excluding the AT3 and MAB files which are effectively general purpose music formats, comes to 1.14 GB for 4211 items totaling 158:50:29. There are a lot of duplicates in there, because for a lot of these items it's more trouble to hunt down a replacement copy than it is to store a backup.
The catch, of course, is that it's all retro videogame music from bleep to bloop.
I mostly use mpv to play local music nowadays. (Most of the music I play is streamed using a Navidrome server with Feishin as the frontend.) Back when I did use a proper audio player on Linux, Harmonoid was my go-to.