you are already specifying it is CLASS B ===> 16 bits for NETWORK part
Now, you are using 12 bits for SUBNET ====> 12 bits for SUBNET part
Total IP address have 32 bits
therefore No.of Host bits = 32 - ( 16 + 12) = 32-28 = 4
with that 4 you can represent 16 addresses
but first address ( put all zero's in host bits ) and last address ( put all one's in host bits ) used for NETWORK IP ADDRESS and BROADCAST IP ADDRESS respectively.
there only 14 hosts possible per subnet