As I understand it, the Apple Hypervisor Kit only does the basics, so underlying CPU & I/O performance should be similar between VMWare, Parallels, UTM and others, but how well they implement things like graphics acceleration and MacOS integration could vary. Parallels has usually been a bit ahead of the game on speed & integration, with VMWare playing close catch-up & arguably being a bit more stable (and better license terms for the Pro version).
Sort of — macOS 10.10 Yosemite introduced Hypervisor.framework, and that's mostly "just" the CPU part of virtualization. Parallels, VMware, etc. used to have their own hypervisor each, but I believe ARM Macs (maybe also Intel Macs at this point) mandate that you use Apple's implementation, so CPU performance will indeed be identical.
But then years later, macOS 11.0 Big Sur introduced Virtualization.framework, which adds stuff like virtual networking, storage, graphics, clipboard sharing, etc., so much so that Apple even provides
sample code for writing your own VM app. (Also, if you virtualize Linux, you can
pass through Rosetta 2 into Linux, letting you run an x86 Linux binary inside the Linux VM through the host macOS's Rosetta.)
But, not for Windows. And, vice versa, Parallels doesn't really offer that stuff for macOS. And on Windows, VMware integration and especially Parallels integration goes deeper, with things like file drag & drop, trackpad gestures, etc.
It seems, though, that Apple is on a path to make virtualized macOS and Linux nicer and nicer, and Parallels and VMware may ultimately decide to use more and more of Apple's built-in stuff and figure out what value they can add on top.