My opinion is that we will see a big improvement in Mac GPU performance but less configurability or variability. The M1 brings entry-level GPU performance to that of contemporary entry-level dedicated GPUs, which is utterly ridiculous if you consider how little power these chips consume. We should expect similar relative improvements for future variants of these chips.
At the same time, since these are chips made by Apple, there will be less variants of them. Now, for example, you can buy the 16" with two different GPU chips (5300M/5500M or 5600M), where 5300M and 5500M are further differentiated by the number of enabled shaders/clocks. I expect the M2 (or whatever it will be called) to offer more streamlined configurations. We will get the 16" with whatever many shader cores the chip offers and that's about it. Maybe there will be two variants: one with couple shader cores disabled for binning...