But I think the reason Mumble has a reputation for poor audio quality is a lot of cheap Mumble server providers. With Mumble, the maximum audio bitrate is set by the server, and many servers set it low to save money on bandwidth, which sometimes ends up being as low as telephone quality. I used to run a Mumble server and had the bitrate set to maximum (I think 128 kbit/sec?), which is absolute overkill for simple voice communications, but it's not like a single server uses much bandwidth. I ran it on a Raspberry Pi for a couple years.
As for latency, not sure how low you expect latency to be. I haven't measured Discord's latency, but I can't imagine it being much more than 100 ms.
Hosting on a cheap VPS in the same country as me, I get 20ms latency, which again is a noticeable difference (the average person can detect latency over 50ms, for example, audio and video can be out of sync by 30ms and it's pretty much undetectable during playback).
The USA is around 2,500 miles wide (+/- 200 miles depending on latitude), which is around 4000 km. If you live on the west coast and connect to a server on the east coast, you're looking at 50 ms minimum based on laws of physics alone. Real world latency is usually 60-70 ms.
Of course, that's a worst case scenario of connecting between the coasts. Honestly I don't know what Discord's voice latency is. I suppose I could measure it, but it's already low enough to be good.
Besides, it's a voice chat service. Even if latency was a ridiculous 200 ms, how would you even notice, other than when you hear your own voice coming from someone else's microphone?