IPv6 Subnet Cheat Sheet

IPv6 addressing works very differently from IPv4 — the numbers are incomprehensibly large, there is no subnet mask in dotted decimal format, and the prefix lengths you encounter in real networks follow a completely different pattern. This chart brings all of that information together in one place so you can quickly look up any prefix or address type without having to calculate or memorise anything.

The page has three tables: the main prefix length reference covering the most important IPv6 prefixes from /0 to /128, an address types table covering every category of IPv6 address and where it is used, and a well-known multicast addresses table for quick reference when reading router configurations.

Here is a plain English explanation of every column in the prefix length table:

1. →Prefix — The CIDR notation prefix length from /0 to /128. Works the same as IPv4 — the number tells you how many of the 128 bits identify the network. The remaining bits identify the device. Unlike IPv4 there is no dotted decimal subnet mask in IPv6 — only the prefix notation is used.
2. →Total Addresses — The number of addresses in the block expressed as a power of 2. For a /64 this is 2⁶⁴ — approximately 18.4 quintillion addresses. These numbers are too large to write out in full so the chart uses power notation throughout.
3. →Number of /64 Subnets — How many standard /64 sized LAN segments fit inside this prefix. This is the most useful column for planning — it tells you how many separate networks you can create from a given allocation. For prefixes larger than /64 (meaning smaller blocks like /96 or /128) this column shows N/A because the block is too small to contain a full /64.
4. →Assigned To — The typical entity or layer in the address allocation hierarchy that holds this prefix size. The chain runs from IANA → Regional Internet Registry → ISP → Customer → LAN segment.
5. →Typical Use / Description — A plain English description of what this prefix is used for in real networks and what you are likely to encounter it for in practice.

In IPv4 you carefully choose the smallest subnet that fits your devices to conserve address space. In IPv6 the design philosophy is the opposite — always use /64 for a LAN segment, no matter how few devices are connected. This is not wasteful — it is intentional, and here is why it matters:

1. 1SLAAC requires /64. Stateless Address Autoconfiguration (SLAAC) is the mechanism that lets IPv6 devices automatically generate their own address without needing a DHCP server. It works by combining the /64 network prefix with a 64-bit Interface Identifier. This only works if exactly 64 bits are available for the interface ID. Using a prefix longer than /64 (like /80 or /96) breaks SLAAC entirely.
2. 2There is no shortage of /64 subnets. Even a typical home broadband /56 allocation contains 256 individual /64 subnets. A business /48 contains 65,536. You will never run out of /64 subnets to assign.
3. 3/64 is the planning unit for IPv6 networks. When designing an IPv6 network, you plan by counting how many /64 subnets you need — one per VLAN, one per LAN segment, one per point-to-point link (though /127 is also used for P2P). Then you choose the parent prefix that contains enough /64 subnets for your design plus room to grow.

If you have been assigned a prefix and want to know how many /64 subnets it contains, the maths is simple:

1. 1Subtract your prefix length from 64. For a /48: 64 − 48 = 16.
2. 2Raise 2 to that power: 2¹⁶ = 65,536 /64 subnets inside your /48.
3. 3If your prefix is already /64 the answer is 1. If your prefix is larger than /64 (meaning a smaller block) the answer is N/A — the block is too small for a /64.

How many /64 subnets fit inside common customer allocations

IPv6 has many address types but in day-to-day network work you will mostly deal with just three. Understanding when each one appears and what it does will help you read router outputs and troubleshoot far more quickly.

1. GUA Global Unicast Address — your public internet address. Starts with 2 or 3 (falls within 2000::/3). This is what your ISP assigns to you and what the rest of the internet sees when you send traffic. Every device on your network can have a Global Unicast address directly — unlike IPv4 where NAT was used to share one public address across many devices. You configure these on interfaces that need to reach or be reached from the internet.
2. LLA Link-Local Address — automatic, local-only, always present. Starts with fe80::. Every single IPv6 interface automatically generates a Link-Local address the moment it comes online — no configuration needed. It only works on the directly connected segment and is never forwarded by a router. Link-Locals are used for router discovery, neighbour discovery (the IPv6 replacement for ARP), and OSPF neighbour relationships. When you see fe80:: in a router config or troubleshooting output this is what it is.
3. ULA Unique Local Address — your private internal address. Starts with fd (falls within fd00::/8). This is the IPv6 equivalent of 192.168.x.x in IPv4 — use it for internal services that should never be accessible from the internet. Unlike Link-Local, Unique Local addresses are routable within your organisation across multiple sites. The fd prefix includes a 40-bit random site identifier that makes it very unlikely two separate organisations using Unique Local addresses will accidentally overlap.

An IPv6 prefix length works exactly like CIDR notation in IPv4. Written as a slash followed by a number from 0 to 128 — for example /64 — it tells you how many of the 128 bits in the address identify the network. The remaining bits identify the specific device. A /64 uses 64 bits for the network and leaves 64 bits for the host, giving 2⁶⁴ — over 18 quintillion — possible device addresses within that single subnet. Unlike IPv4 there is no equivalent to the dotted decimal subnet mask in IPv6. Only prefix notation is used.

The /64 prefix is the standard for a LAN segment because it is required for SLAAC — Stateless Address Autoconfiguration. SLAAC lets devices automatically generate their own IPv6 address by combining the /64 network prefix advertised by the router with a 64-bit Interface Identifier generated from the device's MAC address or a random value. This process only works correctly when exactly 64 bits are available for the host portion. Using any prefix longer than /64 (such as /80 or /96) removes bits from the interface ID space and breaks SLAAC entirely. Even if you only have two devices on a segment, you still assign a /64.

Most ISPs assign a /56 to home broadband customers, which contains 256 individual /64 subnets — more than enough to give every room, guest network, and IoT device its own isolated /64. Business customers typically receive a /48, containing 65,536 /64 subnets — enough for a large organisation to give every floor of every building its own dedicated network segment without ever running short. Smaller business allocations of /52 (4,096 /64s) are also common. ISPs themselves receive /32 allocations from Regional Internet Registries and subdivide them for customers.

No — IPv6 has no broadcast address at all. In IPv4 the last address of every subnet was reserved for broadcast, and sending to it delivered the packet to every device on the segment. IPv6 replaces broadcast entirely with Multicast. Instead of flooding every device, multicast sends packets only to devices that have specifically joined a multicast group. This is far more efficient, especially on large segments. Because there is no broadcast address in IPv6, the last address in any IPv6 block is fully usable and can be assigned to a device — unlike IPv4 where it was always reserved.

A Link-Local address starts with fe80:: and only works on the directly connected network segment. It never crosses a router, is automatically assigned to every interface when it comes online, and is used by the network itself for tasks like router discovery and address resolution. You cannot use a Link-Local address to reach something on a different network. A Global Unicast address starts with 2 or 3 and is a public internet address — routable anywhere in the world, equivalent to a public IPv4 address. Every device on an IPv6 network has both types simultaneously, which is completely normal.

A Unique Local address starts with fd and is the IPv6 equivalent of the private IPv4 ranges like 192.168.x.x. It is usable inside your organisation across multiple sites and routers, but is never routed on the public internet. Use Unique Local addresses for internal services — databases, file servers, management interfaces — that should never be directly reachable from the internet. The fd prefix includes a randomly generated 40-bit site identifier built into the address, which makes accidental address collisions between two different organisations' private networks extremely unlikely — a problem that regularly occurs with IPv4 private ranges when two companies with 192.168.1.0/24 try to merge their networks.

A /128 in IPv6 means all 128 bits identify the network, leaving zero bits for the host. It represents a single specific device address — the IPv6 equivalent of a /32 host route in IPv4. It is used for the loopback address ::1 (the device talking to itself), for individual host routes in routing tables, for BGP peering addresses where you want a stable address that does not depend on any physical interface, and in firewall rules where a rule should apply to exactly one IP address. When you see a /128 in a routing table it always points to or from a single specific device.

Subtract your prefix length from 64 and raise 2 to that power. For a /48: 64 − 48 = 16, so 2¹⁶ = 65,536 /64 subnets. For a /56: 64 − 56 = 8, so 2⁸ = 256 /64 subnets. For a /60: 64 − 60 = 4, so 2⁴ = 16 /64 subnets. If your prefix is already /64 the answer is 1. If your prefix is larger than /64 — meaning a smaller block like /96 or /128 — the block is too small to contain a full /64, so the answer is not applicable. All of these values are pre-calculated in the chart above so you can look them up directly without doing any maths.