Well, currently the M1 and M2 machines crunch MCM almost as good as a great Wintel machine I have while also using far less power. See my "My setup" post on this Hardware Chat Room forum for all the details.

My original post was 2.5 years ago and it's possible that Apple's emulation layer has been improved because I no longer see the RAM issue (knock on wood).

Ultimately it seems that unless the M* machines would crunch much, much faster without the emulation layer, there's really no point in supporting M* natively.

Probably a bigger, and different (and forever has been asked), question is if WCG (or some if its Projects as one-offs) will ever utilize the GPU cores.

I think it has only positiv effects if the binary could be compiled for nativ arm / m1 / m2 processors. I would no expect to change the code a lot tor port it to arm 64-bit.
I would even try to help compiling these binaries if you don't have access to a cross compiler or nativ m2 machines.
I also think GPU support would also be very helpful. There has to be some sort of parallelism possible otherwise the whole project wouldn't work right?
Probably there's a lack of developer writing code for Apple Metal (gpu) or at least cross platform OpenCL code for gpu's on different machines, including the m1/2 gpu.

As today, my mini with M1 works, my air with M3 does not work.

Has Apple Silicon native WU started to be distributed to the Word Community Grid project ?

Well, here we are in 2025 and WCG runs as an "Intel" process meaning it has to be translated ( via Rosetta2 ) from Intel to ARM64 instructions on my M2 Mac mini. Apple's ARM64 Macs have been available since November 2020, so most mac WCG users will be running slower computations. It's surprising WCG hasn't updated ( to Universal so it can run native on either ARM64 or the older Intel ) in over 4 years. Is there any discussion or work to deliver this or, even better, an ETA? Is the source public, maybe that would help speed the transition to current Mac ARM64 support.