Like IPv4, IPv6 addresses can be split into network and host parts using subnet masks.
IPv4 has shown that sometimes it would be nice, if more than one IP address can be assigned to an interface, each for a different purpose (aliases, multi-cast). To remain extensible in the future, IPv6 is going further and allows more than one IPv6 address to be assigned to an interface. There is currently no limit defined by an RFC, only in the implementation of the IPv6 stack (to prevent DoS attacks).
Using this large number of bits for addresses, IPv6 defines address types based on some leading bits, which are hopefully never going to be broken in the future (unlike IPv4 today and the history of class A, B, and C).
Also the number of bits are separated into a network part (upper 64 bits) and a host part (lower 64 bits), to facilitate auto-configuration.