IPv6 Address Types

The Prefix Tells You the Type

In IPv6, an address's purpose is encoded in its leading bits, so the start of the address announces what kind it is. This is different from IPv4, where you had to know the configured ranges. Once you recognise a handful of prefixes, classification is instant. The four types that matter most:

This page assumes you know the basic IPv6 address format — the eight hex groups and the :: shorthand. Here the focus is the taxonomy: what each type does and when you will see it.

Global Unicast: the Public Address

A global unicast address is the IPv6 equivalent of a public IPv4 address — unique across the whole internet and routable everywhere. Current allocations come from 2000::/3, which in practice means an address starting with a 2 or 3, such as 2001:db8::1. This is the address a server presents to the world and the one your device uses to reach external sites over IPv6.

A defining feature of IPv6 is that there is so much address space that public addresses are abundant — every device can have its own genuinely global address, with no shortage forcing the address-sharing that NAT provides in IPv4. That abundance is part of why IPv6 leans on multiple address types rather than translation: a device simply holds a global address for the internet and other types for local jobs, all at once.

Link-Local: Always Present, Never Routed

Every IPv6 interface automatically gives itself a link-local address beginning with fe80, whether or not it has any other address. It works only on the directly connected link and is never forwarded by a router. Far from being a fallback like IPv4's 169.254 addresses, link-local is load-bearing in IPv6: the protocol uses it constantly for essential local machinery — Neighbor Discovery (IPv6's replacement for ARP), finding routers, and bootstrapping the connection before any global address exists.

This is why you always see an fe80 address on an IPv6 interface even when everything is working perfectly. It is not a symptom of a problem; it is the address the device uses to talk to its immediate neighbours and discover the network around it. A device typically has a link-local address and a global one simultaneously, each handling its own scope of communication.

Unique Local: the Private Equivalent

Unique local addresses (ULA) are the IPv6 counterpart to IPv4 private ranges like 192.168.x.x. They live in fc00::/7 and in practice begin with fd. Like private IPv4 addresses, they are used for communication within a site and are not routed on the public internet — ideal for internal services, lab networks, or devices that should never be externally reachable.

ULA improves on IPv4 private addressing in one clever way. IPv4 private ranges are small and identical everywhere, so when two networks merge — a company acquisition, two homes joined by a VPN — their 192.168.1.0 subnets collide. ULA addresses include a large random segment, making each site's prefix globally unique in practice even though it is never routed publicly. Two ULA networks can be joined without renumbering, because the odds of their random prefixes matching are vanishingly small. It is private addressing without the collision headache.

Multicast — and the Broadcast That Isn't

The most striking structural change from IPv4 is what is missing: IPv6 has no broadcast address. The job broadcast did — reaching many hosts at once — is handled entirely by multicast, which sends only to devices that have joined a specific group rather than interrupting everyone. Multicast addresses begin with ff, and IPv6 defines well-known groups for the tasks IPv4 broadcast used to cover:

• ff02::1 — the all-nodes group, reaching every device on the link (the closest thing to old broadcast).
• ff02::2 — the all-routers group, reaching the routers on the link.

Removing broadcast was deliberate. Broadcast forced every host to process traffic whether relevant or not, creating needless noise; multicast targets only the interested, which is more efficient and scales better. So when an IPv6 device needs to "ask the whole link" something, it sends to the all-nodes multicast group rather than shouting a broadcast. The shift mirrors the broader IPv6 philosophy of doing the same jobs more precisely — and it is why, unlike every IPv4 subnet, an IPv6 network reserves no broadcast address at all. There is also anycast, where one address is assigned to several locations and traffic goes to the nearest, but the four types above are the ones you will meet daily.

Frequently Asked Questions

What are the main IPv6 address types?

Global unicast (publicly routable, starting 2000::/3), link-local (fe80::/10, single-link only), unique local (fc00::/7, the private equivalent), and multicast (ff00::/8, one-to-many). Unlike IPv4, IPv6 has no broadcast — multicast handles those jobs.

What is an IPv6 link-local address?

An address beginning with fe80, automatically assigned to every IPv6 interface and never routed beyond the local link. IPv6 relies on it for neighbor discovery and finding routers, so every interface always has one even with no other address.

What is the IPv6 equivalent of a private IP address?

Unique local addresses (ULA) in fc00::/7, in practice starting with fd. They are the counterpart to IPv4 private ranges, used internally and not routed publicly — but designed to be globally unique to avoid collisions when networks merge.

Why doesn't IPv6 have a broadcast address?

It was removed for inefficiency — broadcast forced every device to process traffic regardless of interest. IPv6 uses multicast instead, sending only to devices that joined a group. Groups like all-nodes (ff02::1) and all-routers (ff02::2) cover the old broadcast cases.

What does a global unicast IPv6 address look like?

The public, internet-routable type, with current allocations starting 2000::/3 — addresses beginning with 2 or 3, like 2001:db8::1. It is the equivalent of a public IPv4 address, and a device can hold one alongside a link-local address at the same time.