When a vendor like Analog Devices offers a DAC family with nearly identical parts like the AD5671/AD5721 are both devices the same die but rated into different accuracy classes like CPUs? Is the 12-bit version just the dies where the high precision part of the R/2R ladder failed testing so they disabled the lowest 4 bits with a fuse or something?
They are usually the same die.
Whereas performance binning is often the reason to have parts of two difference grades, there is another, and to some people surprising, reason for different grades.
With a 12/16 bit split, you may well find that if a part works at all, it works to 12 bits 'straight off the line', and needs no more than a functional test. The 16 bit parts will need more testing, and any rejects from that will indeed find their way back to the 12 bit line.
Testing to high performance costs money. You need to use accurate equipment, and the testing will take longer. Not only linearly longer because there are more tests, but often exponentially longer due to the waiting times for readings to settle, or temperatures to stabilise.
So part of the difference in selling price is the difference in performance. But part is also the cost of production, the fact that the 12 bit parts don't tie up a large amount of expensive test gear.