// FREE IT TOOLS

IPv6 Subnet Calculator

Compress, expand and break down any IPv6 address — and see exactly what kind of address it is.

// IPv6 ADDRESS OR PREFIX

Enter an IPv6 address with an optional prefix length

Documentation

Reserved for examples and documentation. It must never appear on a real network, so seeing it in traffic is a finding.

Canonical form

2001:db8:abcd:12::

Fully expanded

2001:0db8:abcd:0012:0000:0000:0000:0000

Network address

2001:db8:abcd:12::

Prefix length

/64

// ADDRESS RANGE

First address

2001:db8:abcd:12::1

Last address

2001:db8:abcd:12:ffff:ffff:ffff:ffff

Total addresses

18,446,744,073,709,551,616

Subnet bits

64 (prefix)

Interface ID bits

64 (host)

// EUI-64 FROM A MAC ADDRESS

MAC address

Interface identifier

200:5eff:fe00:5301

Link-local address

fe80::200:5eff:fe00:5301

Need the vendor behind a MAC? Use the MAC Address Lookup to identify the manufacturer from the IEEE OUI registry.

// PREFIX LENGTH REFERENCE

Prefix	Typical use	/64 subnets	Notes
/32	ISP allocation	4,294,967,296	The smallest block a regional registry typically hands an ISP. You will rarely see one handed to an organisation directly.
/48	Site	65,536	The conventional allocation for one site or organisation: 65,536 /64 subnets, far more than most networks will ever consume.
/56	Residential / small site	256	What many ISPs delegate to a home or small business connection — 256 /64 subnets, enough to segment a home network several times over.
/64	Single LAN	1	The standard size for one subnet. SLAAC requires a /64, so a LAN that must support stateless autoconfiguration cannot be any other size.
/127	Point-to-point link	—	RFC 6164 recommends /127 for the link between two routers. It removes the subnet-router anycast ambiguity and the ping-pong risk of a /64 on a point-to-point segment.
/128	Single host	—	One specific address rather than a subnet — used for loopbacks and static host routes.

IPv6 Addressing, Briefly

An IPv6 address is 128 bits, written as eight groups of four hexadecimal digits separated by colons. That is four times the length of an IPv4 address, which is why the notation needs compression rules — and why two addresses that look completely different can be the same address.

The first 64 bits are conventionally the network portion and the last 64 bits the interface identifier. This split is not enforced by the protocol, but so much of IPv6 assumes it — SLAAC most of all — that treating it as the default is the practical choice. It also means the numbers involved are genuinely large: a single /64 holds 18,446,744,073,709,551,616 addresses.

That scale is the mental shift IPv6 requires. IPv4 subnetting is about conserving a scarce resource, which is why variable-length subnet masks and careful address planning matter so much. IPv6 subnetting is about counting subnets, not addresses: the question is never “how do I fit this in 30 usable addresses” but “how many /64s do I need, and do I have enough bits to number them”.

The Compression Rules That Trip People Up

RFC 5952 exists so that one address has exactly one canonical written form — without it, a log correlation or an access-list comparison becomes unreliable because the same address renders two different ways. The rules are narrower than most people assume:

Lowercase only

Hex digits are always lowercase — 2001:db8::, never 2001:DB8::.

Strip leading zeros per group

A group is written without leading zeros: 0db8 becomes db8. A group that is all zeros becomes a single 0 — unless it is part of a compressed run.

Compress only runs of two or more

:: must stand in for at least two zero groups. A lone zero group is left as 0, which is why 1:0:2:3:4:5:6:7 keeps its zero.

Longest run wins

If there are several zero runs, :: replaces the longest one — 1:0:0:2:0:0:0:3 compresses to 1:0:0:2::3, not 1::2:0:0:0:3.

Leftmost wins a tie

When two runs are the same length, the leftmost one is compressed: 1:0:0:2:3:0:0:4 becomes 1::2:3:0:0:4.

The rule people most often get wrong is the first of the tie-breakers. Collapsing the longest run is not the same as collapsing the first run, and a calculator that gets it wrong will produce a string that is valid but not canonical — which is exactly the kind of difference that makes two log lines fail to match.

Reading the Address You Have

The prefix tells you what an address is for before you look at anything else. These are the ones you will meet most often.

Prefix	Purpose	Routed?
2000::/3	Global unicast — the public internet	Yes
fc00::/7	Unique local (ULA) — private internal networks	Internal only
fe80::/10	Link-local — automatic on every interface	Never
ff00::/8	Multicast — replaces broadcast entirely	Per scope
::1/128	Loopback	Never
2001:db8::/32	Documentation only	Should never appear
::ffff:0:0/96	IPv4-mapped — an IPv4 address wearing IPv6 notation	Unwrapped to IPv4

If you are auditing a network, seeing 2001:db8::/32 in live traffic is worth investigating: it is reserved for documentation, so it should never be configured on a real device. And a globally routable address on an internal interface — rather than a ULA — is often a sign that the address plan was never finished.

Planning a /48, /56 or /64

A common mistake is to plan an IPv6 network the way you would plan an IPv4 one — carving out small subnets to conserve addresses. There is no reason to. The bits between the delegated prefix and the /64 boundary are your subnet field, and each of them doubles the number of LANs you can number:

Delegated	Subnet bits	/64 LANs available
/48	16	65,536
/52	12	4,096
/56	8	256
/60	4	16
/64	0	1

A sensible addressing plan numbers the subnet field meaningfully — for example, reserving a block for infrastructure links and another for user VLANs, and mirroring the last octet of an existing IPv4 plan into the subnet field to keep the mental mapping intact while both protocols run side by side. What you should not do is subdivide below /64 on a LAN, because that breaks SLAAC.

For IPv4 work, the IPv4 Subnet Calculator covers masks, broadcast addresses and usable host ranges, and the CIDR Cheat Sheet is the quick reference from /0 to /32.

Frequently Asked Questions

Why is a /64 the standard size for an IPv6 subnet?: Because stateless address autoconfiguration (SLAAC) requires it. An interface identifier is 64 bits, so a LAN that lets devices configure themselves from router advertisements must leave exactly 64 bits for that identifier. A /64 also gives 18.4 quintillion addresses per subnet, which is why IPv6 subnetting is about counting subnets rather than conserving addresses — you almost never need to make a subnet smaller than /64.
How many /64 subnets are in a /48?: 65,536. A /48 leaves 16 bits for subnetting (64 minus 48), so there are 2^16 subnets of /64. A /56 leaves 8 bits, giving 256 /64s, which is why that is the size many ISPs delegate to a home or small business connection. The calculation is always 2 raised to the power of 64 minus the prefix length.
How do I compress an IPv6 address correctly?: RFC 5952 gives the rules, and they are stricter than most people assume. Write hex in lowercase, strip leading zeros from each group, and replace the longest run of consecutive zero groups with :: — but only if that run is two groups or longer, and the leftmost run wins if two runs are the same length. A single zero group is never compressed. This tool applies exactly those rules, which is why ::1 stays ::1 and 1:0:2:3:4:5:6:7 keeps its lone zero.
Does IPv6 have a broadcast address?: No. IPv6 has no broadcast at all — multicast replaces it entirely. That removes the IPv4 concepts of a network address and a broadcast address that cannot be assigned to a host. The all-zeros interface identifier is reserved as the subnet-router anycast address, so in practice the first ordinary host on a subnet is the network address plus one, which is what this calculator shows.
What is EUI-64 and is it a privacy problem?: EUI-64 builds a 64-bit interface identifier from a device's 48-bit MAC address by inserting ff:fe in the middle and flipping the universal/local bit. That means the address contains the hardware identifier, so a device keeps the same interface ID on every network it joins — which makes it trivial to track that device across networks and to correlate its traffic. RFC 4941 defined privacy extensions, where the interface identifier is randomised and periodically replaced, and that is what most modern operating systems use by default.
What is the difference between link-local, unique local and global unicast?: Link-local addresses live in fe80::/10 and are valid only on a single link — never routed. Every IPv6 interface has one automatically, and they are how routers and neighbour discovery are reached. Unique local addresses live in fc00::/7 and are the IPv6 equivalent of RFC 1918 private space: routable inside your organisation but not on the internet. Global unicast addresses live in 2000::/3 and are the publicly routable addresses. It is normal and expected for one interface to hold all three at once.
Do I still need NAT with IPv6?: No, and you should not use it. NAT exists in IPv4 to work around address exhaustion; IPv6 has enough addresses that every device can have a globally routable one. Address translation also breaks the end-to-end connectivity that IPv6 features such as IPsec and peer-to-peer connections rely on. A stateful firewall gives you the same inbound protection NAT appeared to provide, without the translation — the security you want comes from the firewall policy, not from the address shortage.
What does a /127 mean on a router-to-router link?: RFC 6164 recommends /127 for links between two routers. The older advice was to use a /64 everywhere for consistency, but a /64 on a point-to-point link creates two known problems: the subnet-router anycast address is ambiguous, and it opens a ping-pong risk where the two routers can bounce traffic between themselves. A /127 assigns exactly the two addresses the link needs and removes both issues.

// NETWORK INFRASTRUCTURE

Planning an IPv6 rollout?

DMC IT Services provides network design, IPv6 migration planning, and infrastructure support for SMBs across London, Cambridge, Hertfordshire, and Bedfordshire.

Talk to an Engineer

← Back to all free tools

Prefix

Typical use

/64 subnets

Notes

/32

ISP allocation

4,294,967,296

The smallest block a regional registry typically hands an ISP. You will rarely see one handed to an organisation directly.

/48

Site

65,536

The conventional allocation for one site or organisation: 65,536 /64 subnets, far more than most networks will ever consume.

/56

Residential / small site

256

What many ISPs delegate to a home or small business connection — 256 /64 subnets, enough to segment a home network several times over.

/64

Single LAN

The standard size for one subnet. SLAAC requires a /64, so a LAN that must support stateless autoconfiguration cannot be any other size.

/127

Point-to-point link

—

RFC 6164 recommends /127 for the link between two routers. It removes the subnet-router anycast ambiguity and the ping-pong risk of a /64 on a point-to-point segment.

/128

Single host

—

One specific address rather than a subnet — used for loopbacks and static host routes.

IPv6 Addressing, Briefly

The Compression Rules That Trip People Up

Lowercase only

Hex digits are always lowercase — 2001:db8::, never 2001:DB8::.

Strip leading zeros per group

A group is written without leading zeros: 0db8 becomes db8. A group that is all zeros becomes a single 0 — unless it is part of a compressed run.

Compress only runs of two or more

:: must stand in for at least two zero groups. A lone zero group is left as 0, which is why 1:0:2:3:4:5:6:7 keeps its zero.

Longest run wins

If there are several zero runs, :: replaces the longest one — 1:0:0:2:0:0:0:3 compresses to 1:0:0:2::3, not 1::2:0:0:0:3.

Leftmost wins a tie

When two runs are the same length, the leftmost one is compressed: 1:0:0:2:3:0:0:4 becomes 1::2:3:0:0:4.

Reading the Address You Have

The prefix tells you what an address is for before you look at anything else. These are the ones you will meet most often.

Prefix	Purpose	Routed?
2000::/3	Global unicast — the public internet	Yes
fc00::/7	Unique local (ULA) — private internal networks	Internal only
fe80::/10	Link-local — automatic on every interface	Never
ff00::/8	Multicast — replaces broadcast entirely	Per scope
::1/128	Loopback	Never
2001:db8::/32	Documentation only	Should never appear
::ffff:0:0/96	IPv4-mapped — an IPv4 address wearing IPv6 notation	Unwrapped to IPv4

Planning a /48, /56 or /64

Delegated	Subnet bits	/64 LANs available
/48	16	65,536
/52	12	4,096
/56	8	256
/60	4	16
/64	0	1

For IPv4 work, the IPv4 Subnet Calculator covers masks, broadcast addresses and usable host ranges, and the CIDR Cheat Sheet is the quick reference from /0 to /32.

Frequently Asked Questions

Why is a /64 the standard size for an IPv6 subnet?

Because stateless address autoconfiguration (SLAAC) requires it. An interface identifier is 64 bits, so a LAN that lets devices configure themselves from router advertisements must leave exactly 64 bits for that identifier. A /64 also gives 18.4 quintillion addresses per subnet, which is why IPv6 subnetting is about counting subnets rather than conserving addresses — you almost never need to make a subnet smaller than /64.

How many /64 subnets are in a /48?

65,536. A /48 leaves 16 bits for subnetting (64 minus 48), so there are 2^16 subnets of /64. A /56 leaves 8 bits, giving 256 /64s, which is why that is the size many ISPs delegate to a home or small business connection. The calculation is always 2 raised to the power of 64 minus the prefix length.

How do I compress an IPv6 address correctly?

RFC 5952 gives the rules, and they are stricter than most people assume. Write hex in lowercase, strip leading zeros from each group, and replace the longest run of consecutive zero groups with :: — but only if that run is two groups or longer, and the leftmost run wins if two runs are the same length. A single zero group is never compressed. This tool applies exactly those rules, which is why ::1 stays ::1 and 1:0:2:3:4:5:6:7 keeps its lone zero.

Does IPv6 have a broadcast address?

No. IPv6 has no broadcast at all — multicast replaces it entirely. That removes the IPv4 concepts of a network address and a broadcast address that cannot be assigned to a host. The all-zeros interface identifier is reserved as the subnet-router anycast address, so in practice the first ordinary host on a subnet is the network address plus one, which is what this calculator shows.

What is EUI-64 and is it a privacy problem?

EUI-64 builds a 64-bit interface identifier from a device's 48-bit MAC address by inserting ff:fe in the middle and flipping the universal/local bit. That means the address contains the hardware identifier, so a device keeps the same interface ID on every network it joins — which makes it trivial to track that device across networks and to correlate its traffic. RFC 4941 defined privacy extensions, where the interface identifier is randomised and periodically replaced, and that is what most modern operating systems use by default.

What is the difference between link-local, unique local and global unicast?

Link-local addresses live in fe80::/10 and are valid only on a single link — never routed. Every IPv6 interface has one automatically, and they are how routers and neighbour discovery are reached. Unique local addresses live in fc00::/7 and are the IPv6 equivalent of RFC 1918 private space: routable inside your organisation but not on the internet. Global unicast addresses live in 2000::/3 and are the publicly routable addresses. It is normal and expected for one interface to hold all three at once.

Do I still need NAT with IPv6?

No, and you should not use it. NAT exists in IPv4 to work around address exhaustion; IPv6 has enough addresses that every device can have a globally routable one. Address translation also breaks the end-to-end connectivity that IPv6 features such as IPsec and peer-to-peer connections rely on. A stateful firewall gives you the same inbound protection NAT appeared to provide, without the translation — the security you want comes from the firewall policy, not from the address shortage.

What does a /127 mean on a router-to-router link?

RFC 6164 recommends /127 for links between two routers. The older advice was to use a /64 everywhere for consistency, but a /64 on a point-to-point link creates two known problems: the subnet-router anycast address is ambiguous, and it opens a ping-pong risk where the two routers can bounce traffic between themselves. A /127 assigns exactly the two addresses the link needs and removes both issues.