Geekbench 5 isn't perfect, but it has begun to crop up - there are now a lot of things claiming to be MacPro7,1 that may actually be... MacPro designators are always tricky, because a lot of Hackintoshes use the MacPro IDs, so anything reporting a "wrong" processor is almost certainly a Hackintosh (and there may be Hackintoshes using or reporting Mac Pro CPUs).
The other problem with brand-new machines is Spotlight and friends killing CPU for the first couple of days... The safest thing to do is to take only the HIGHER numbers for just-released machines - it's far more likely that there's a machine reporting an anomalously low score due to some setup process than a high one due to a "golden sample" chip. That said, if I saw 20 scores right in the same range and one 15% higher, I might suspect an overclocked Hackintosh was responsible for the high score.
The 8 core is scoring right around 8000 (multi-core) in the Geekbench browser list - some scores around 6000 are probably affected by Spotlight or other setup processes. One user posted a score over 8400 here in another thread.
The 12 core is right around 12000, with a few scores a little over (a few hundred points).
There are a lot of 16 core Hackintoshes using "wrong chips" (including Ryzens), so I'm not confident that none of the 16-core results showing a Xeon W-3245 are also Hackintoshes - but there are quite a few right under 15000 - at least many of which are probably Mac Pros. There is one score just over 15500, while nothing else breaks 15000 - that could be the first machine that finished its setup process fully - or it could be a slightly overclocked Hackintosh? The user has posted a lot of scores from confirmable Apple devices (and no obvious Hackintoshes), so it's relatively likely to be a Mac Pro.
There's one 24 core posting over 19000, with the rest around 16500 - my guess is that the higher number is what we'll end up seeing, and the two or three lower numbers are Spotlight. It doesn't make sense that the 24 core isn't faster than the 16 - unless there's a bug in Geekbench about large numbers of cores (but where did that higher number come from then)?
No 28 cores in the browser yet, but Marques Brownlee put a score around 21500 on YouTube.