The IPv6 fragment header is 64 bits long. The fragment header is used by an IPv6 source to indicate a packet that exceeds the path maximum transmission unit (MTU) size. Unlike IPv4, which enables intervening devices such as routers to fragment packets that exceed the permitted size for a local link, IPv6 requires the traffic originator to ensure that each packet sent is small enough to traverse the entire link without fragmentation. The packet can then be reassembled at the destination.

The IPv6 fragment header is not 32 bits long. However, the IPv6 fragment header contains a 32bit field called the identification field. The identification field is used to uniquely identify each fragmented packet.

The IPv6 fragment header is neither 20 bytes nor 40 bytes long. A basic IPv4 header without options is 20 bytes long, and a basic IPv6 header without extension headers is 40 bytes long. Although an IPv4 header is shorter than an IPv6 header, it is more complex and contains more fields than an IPv6 header. Several fields that exist in an IPv4 header, such as the Header Checksum field and the Fragment Offset field, do not exist in an IPv6 header. Because many protocols at the Data Link and Transport layers contain mechanisms to verify the integrity of the packet, IPv6 does not contain a redundant method to calculate checksum values.

The IPv6 fragment header is not 1,280 bytes long. The default IPv6 MTU size is 1,280 bytes. IPv6 requires that each device have an MTU of 1,280 bytes or greater.