I have some legacy products at work that use 8051 but nothing new. That's all NXP arms. However I feel like I don't actually understand the modern 8051 landscape very well. I know the core goes into a lot of very modern micros and it's still used a lot, but I don't really know who uses them and why they do, aside from the "no core royalty makes them cheap" argument. Again, what does that cost? I don't know because I'm not plugged into that at all.
An 8051-variant vendor doesn't need to pay Intel a royalty for intellectual property. ARM is not a manufacturer; they hold the IP and license it to anyone willing to pony up for it. I'm not sure that such lack of royalty in itself makes the 8051 significantly less expensive than ARM, for example. In fact, just scan some price lists and you'll see that many well-outfitted ARMs are cheaper than some 8051s.
Seems like most core arguments are mostly a question of familiarity and much less about what the actual differences are.
That's often the case.
But look at it this way: if you've sprung for the Keil 8051 toolset and your applications are not strictly cost-sensitive (that is, you're not selling lots of product per month), why not continue to use the old familiar? It works, you have the tools and as such your development time is reduced.
In these cases, you'd consider an alternate device family if you need more horsepower or peripherals which your preferred family doesn't have.
If you're starting from tabula rasa, then I'd say look at ARMs and skip the 8051.