IPv6 Subnet Calculator

Type an IPv6 prefix to see its range, how many /64s it holds and its reverse zone, then split it or plan sites and VLANs on 4-bit nibble boundaries.

RFC 4291 and RFC 5952 arithmetic, exact to the last address · RIR and RIPE-690 sizes · runs in your browser

Type an IPv6 prefix, an address with its prefix length, or a range with a dash. Drag the line between the bits to change the prefix. Nothing leaves your browser.

Documentation · RFC 3849/48site

65,536 /64 subnets · 1,208,925,819,614,629,174,706,176 addresses

  • There is no broadcast in IPv6. The address with an all-zero interface ID is the Subnet-Router anycast address (RFC 4291), and the top 128 interface IDs of each subnet are kept for anycast (RFC 2526).
  1. 1

    The prefix

    /48 keeps the first 48 bits: 12 hex digits

    address2001:0db8:abcd:0000:0000:0000:0000:0000
  2. 2

    Network = the prefix, then zeros

    network2001:0db8:abcd:0000:0000:0000:0000:00002001:db8:abcd::
  3. 3

    Last address = the prefix, then ones

    IPv6 has no broadcast address

    last2001:0db8:abcd:ffff:ffff:ffff:ffff:ffff2001:db8:abcd:ffff:ffff:ffff:ffff:ffff
  4. 4

    Addresses = 2128 − 48 = 280

    Exactly 1,208,925,819,614,629,174,706,176

  5. 5

    /64 subnets = 264 − 48 = 65,536

How does IPv6 subnetting work?

An IPv6 address is 128 bits: a routing prefix from your provider, a subnet ID you choose for each network, and a 64-bit interface ID each device fills in. Subnetting IPv6 means choosing the subnet ID's bits.

RFC 4291 sets the layout: a global routing prefix, a subnet ID and an interface ID, and every global address has a 64-bit interface ID. So the routing prefix and the subnet ID share the first 64 bits, and one network is a /64. The classic layout, from RFC 3587, is a /48 routing prefix and a 16-bit subnet ID: 65,536 networks per site.

You don't size IPv6 subnets by host count as in IPv4: every LAN is a /64 whatever its size, because stateless address autoconfiguration (SLAAC) needs a 64-bit interface ID (RFC 7421). What you plan is how many subnets you need and how to group them.

Type any prefix in the IPv6 calculator above to see its bits on the ruler, then Split or Plan it. Range → CIDR makes it an IPv6 range calculator: type two addresses with a dash for the fewest prefixes that cover them.

One IPv6 address, three parts2001:db8:abcd:12::1/64

2001:0db8:abcd:0012:0000:0000:0000:0001

  • Global routing prefix48 bits · /0 → /482001:0db8:abcd
  • Subnet ID16 bits · /48 → /640012
  • Interface ID64 bits · /64 → /1280000:0000:0000:0001
The classic site layout (RFC 3587): a /48 from the provider, 16 bits of subnet ID for your networks, 64 bits of interface ID for each device.
The IPv6 Subnet Calculator showing 2001:db8:abcd::/48 on its nibble ruler, with its counts, reverse zone and the Plan tabThe IPv6 Subnet Calculator showing 2001:db8:abcd::/48 on its nibble ruler, with its counts, reverse zone and the Plan tabThe prefix, bit by bitThe ruler marks the usual boundaries: /32 for an ISP, /48 for a site, /64 for a LAN.A new ULA /48A random unique local prefix, made in your browser and never sent.Exact counts65,536 /64 subnets and every address in the block, counted exactly.Its reverse zoneThe ip6.arpa zone where the block's reverse DNS lives, beside its full 32-digit form.
2001:db8:abcd::/48, a site from the documentation prefix: its bits on the nibble ruler, 65,536 /64 subnets counted exactly, its reverse zone, and the Plan tab.

How many /64 subnets are in a /48, /56 or /60?

A /48 holds 65,536 /64 subnets, a /56 holds 256 and a /60 holds 16. A prefix n bits shorter than /64 holds 2ⁿ of them: every 4 bits is one more hex digit and 16 times as many.

The table counts exactly, to the last address. The sizes people meet: a /32 is the smallest block ARIN, the RIPE NCC and APNIC allocate to an ISP, a /48 is a site, a /56 is the smallest prefix RIPE-690 advises ISPs to give a home, and a /64 is one LAN.

As an IPv6 prefix calculator, the tool above counts any length the same way, /0 to /128. Inside a /64 the count is addresses, not subnets: 18,446,744,073,709,551,616 of them. Nobody fills a /64; the space lets devices pick their own addresses without asking, which RFC 7934 recommends networks allow.

Subnets and addresses per prefix length
PrefixTypical use/48s/56s/64 subnetsAddresses
/32an ISP's smallest allocation65,53616,777,2164,294,967,29629679,228,162,514,264,337,593,543,950,336
/364,0961,048,576268,435,4562924,951,760,157,141,521,099,596,496,896
/4025665,53616,777,216288309,485,009,821,345,068,724,781,056
/44164,0961,048,57628419,342,813,113,834,066,795,298,816
/48a site125665,5362801,208,925,819,614,629,174,706,176
/52–164,09627675,557,863,725,914,323,419,136
/56a home (RIPE-690's smallest)–12562724,722,366,482,869,645,213,696
/60––16268295,147,905,179,352,825,856
/64one LAN––126418,446,744,073,709,551,616
Exact counts from this page's engine. A dash: the block is smaller than that size.

What IPv6 prefix will my ISP give me?

Your router receives a prefix from the ISP by DHCPv6 prefix delegation and gives each of its networks a /64 from it. RIPE-690 advises ISPs to give every customer a /48, or businesses a /48 and homes a /56, always on a whole hex digit: /48, /52, /56 or /60.

The chain runs from the registries down. A regional registry allocates an ISP at least a /32 (ARIN, RIPE NCC, APNIC). The ISP delegates a prefix to your router with DHCPv6 prefix delegation (RFC 9915, which replaced RFC 8415 in January 2026), and the router gives each of its LAN interfaces its own /64 from that prefix (RFC 7084, requirement L-2).

A router may hint at the size it wants, rounded up to a whole hex digit, but it must accept what it gets (RFC 7084). Since RFC 9818 (July 2025), a home router must keep the part of the prefix it doesn't use available to other routers behind it.

Few ISPs publish the size on their own pages, so look for the delegated prefix on your router's IPv6 status page and type it in above. RIPE-690 strongly discourages delegating anything longer than a /56, and a single /64 leaves you nothing to split. The IPv6 Test shows whether your connection uses IPv6 at all.

From the registry to your LAN
Regional registryallocates the ISP/32
16,777,216 × /56
ISPdelegates to your router/56
256 × /64
Your routergives each LAN/64
The /56 is an example: the ISP chooses the size, and RIPE-690 advises a /48, or a /56 for homes.

How do I plan IPv6 subnets on nibble boundaries?

Give each level of your network whole hex digits: one digit for up to 16 sites, two for up to 256 VLANs. Every block then starts and ends on a digit, reads cleanly, and maps to exactly one ip6.arpa zone for reverse DNS.

Reverse DNS names an IPv6 address digit by digit, in reverse, under ip6.arpa (RFC 3596), so a reverse zone can only be cut at a hex digit. RIPE-690 asks ISPs to delegate prefixes on 4-bit (nibble) boundaries to align with those zones, and ARIN makes its allocations on nibble boundaries.

Plan from the top: name each level, count what you need now, round each level up to whole digits and keep a digit spare above the first level. The figure is the Plan tab's answer for a /48 with 3 sites and 20 VLANs each: the sites take one digit, the VLANs two, and one digit stays in reserve.

A prefix off a nibble boundary still routes (RFC 7608 requires forwarding on any length), but its reverse DNS spans several zones: a /57 needs 8, a /58 needs 4.

A /48 planned on whole hex digits2001:db8:abcd::/48

2001:0db8:abcd:0213:0000:0000:0000:0000

  • Your /48
  • Reserve
  • Sites 3
  • VLANs 20
  • Interface ID

Reverse zone of site 3, VLAN 203.1.2.0.d.c.b.a.8.b.d.0.1.0.0.2.ip6.arpa

Every level on whole digits: site 3, VLAN 20 is 2001:db8:abcd:213::/64, and its reverse zone is a single ip6.arpa name.

Which IPv6 prefix lengths should I use?

A /64 for every LAN, a /127 between two routers, a /128 for a loopback, and a /48 or /56 for a whole site. A LAN with a prefix longer than /64 loses SLAAC.

SLAAC needs a /64 (RFC 4291, RFC 7421). On a link between two routers, RFC 6164 made /127 standard, the IPv6 version of an IPv4 /31, and routers must support it. Prefixes longer than /64 are valid: RFC 7608 requires routers to forward on any length. They only lose SLAAC, so they suit router links and servers numbered by hand, not networks of phones and laptops.

Prefix lengths by use
UseLengthWhy
A LAN (phones, laptops, servers)/64SLAAC needs a 64-bit interface ID (RFC 4291, RFC 7421)
A link between two routers/127The IPv6 /31; routers must support it (RFC 6164)
A loopback or one host/128A single address
A home/56256 LANs; RIPE-690's smallest for a customer
A site or business/4865,536 LANs; RIPE-690's advice for every customer
An ISP/32The registries' smallest allocation (ARIN, RIPE NCC, APNIC)
Lengths longer than /64 are valid everywhere (RFC 7608); only SLAAC needs exactly /64.

What is a ULA, and how do I generate one?

A unique local address (ULA) prefix is fd followed by a random 40-bit Global ID: a /48 for use inside your site that isn't routed on the Internet. The New ULA /48 chip above is a ULA generator: the random bits come from your browser and are never sent.

RFC 4193 reserves fc00::/7. The next bit, L, is 1 for a prefix you assign yourself, so in practice every ULA starts with fd. The 40-bit Global ID must be pseudo-random, never sequential or a well-known number, so that two networks joined later almost never clash.

RFC 4193 suggests making the Global ID from SHA-1 over the time and a hardware identifier, but any good random source meets the requirement; this page uses your browser's cryptographic random generator. The odds of a clash stay small: the figure applies the RFC's own formula.

A ULA is not a firewall or NAT: it is simply not routed on the Internet. Use one beside your global prefix for internal services that should keep their addresses when the ISP's prefix changes.

A ULA /48, part by partfd5b:3c91:e72::/48 · example
  1. fdfc00::/7 + L = 18 bits
  2. 5b:3c91:0e72Global ID, random40 bits
  3. 0000–ffffSubnet ID16 bits
  4. …Interface ID64 bits
Chance two Global IDs clash (RFC 4193 §3.2.3)
Networks joinedChance of a clash
10≈ 1 in 22,000,000,000
100≈ 1 in 220,000,000
1,000≈ 1 in 2,200,000
10,000≈ 1 in 22,000
The example's Global ID is fixed for this figure; the chip above draws a new random one each time.

IPv4 vs IPv6 subnetting: what changes?

In IPv4 you size each subnet by its host count and lose two addresses to the network and broadcast; in IPv6 every LAN is a /64, there is no broadcast, and you plan how many subnets rather than how many hosts.

What carries over: CIDR notation, splitting a block into equal halves, and summarizing neighbours into one prefix. The Subnet Calculator does both families; this page starts on IPv6 and adds the Plan tab.

IPv4 and IPv6 subnetting side by side
IPv4IPv6
Address size32 bits128 bits
Written as192.168.1.0/24 or a mask, 255.255.255.02001:db8:abcd:12::/64, prefix length only
BroadcastThe last address of each subnetNone: multicast (ff02::1) does that job
Usable addresses2ⁿ − 2All but the Subnet-Router anycast (interface ID 0) and the top 128 interface IDs (RFC 2526)
Subnet sizeSized to its hosts (VLSM)Always /64 for a LAN
Private ranges10/8, 172.16/12, 192.168/16 (RFC 1918)fd00::/8 unique local, a random /48 per site (RFC 4193)
Reverse DNSin-addr.arpa, one label per octetip6.arpa, one label per hex digit
Sources: RFC 4291 (no broadcast, the anycast address), RFC 2526, RFC 4193, RFC 1918, RFC 3596.

IPv6 subnetting questions

How many /64 subnets are in a /48?

65,536. A /48 leaves 16 bits before the 64-bit interface ID, and 2^16 = 65,536. A /56 holds 256 /64s (8 bits), a /60 holds 16 (4 bits), and a /52 holds 4,096. Each step of 4 bits is one more hex digit of subnet ID, which is why plans keep to /48, /52, /56, /60 and /64.

How many addresses are in a /64?

18,446,744,073,709,551,616, which is 2^64. A /64 is the size of one LAN: the last 64 bits are the interface ID each device picks for itself (RFC 4291). Nobody fills a /64; it is sized so devices can choose addresses on their own without colliding.

Why is every IPv6 LAN a /64?

Stateless address autoconfiguration (SLAAC), the way most devices give themselves an address, builds a 64-bit interface ID, so it needs a /64 (RFC 4291, RFC 7421). A LAN with a longer prefix can't use SLAAC, only DHCPv6 or addresses set by hand. Router-to-router links are the exception: /127 is standard there (RFC 6164).

Does IPv6 have a broadcast address?

No. IPv6 has no broadcast; multicast does that job, for example ff02::1 for every node on the link (RFC 4291). So the last address of a block is an ordinary address, and you don't subtract two as in IPv4. The address with an all-zero interface ID is the Subnet-Router anycast address.

What is the difference between a /48 and a /56?

Eight bits, or 256 times the space. A /48 holds 65,536 LANs of /64; a /56 holds 256. RIPE-690 advises ISPs to give every customer a /48, or a /48 to businesses and a /56 to homes, and strongly discourages anything longer than a /56, so customers can split their prefix without renumbering.

Can I use a /127 between routers?

Yes. RFC 6164 made /127 the standard for point-to-point links between routers, the IPv6 version of an IPv4 /31, and routers must support it. It replaced the older advice against it (RFC 3627, now Historic). When several /127s come out of one /64, leave out the all-zero interface ID and the top 128 values.

What is a nibble boundary?

A prefix length that is a multiple of 4, such as /48, /52 or /56. Each nibble is one hex digit, so a block on a nibble boundary starts and ends on whole digits (2001:db8:abcd:1200::/56 rather than a /57), reads more easily, and maps to exactly one ip6.arpa zone for reverse DNS. Address plans keep each level to whole nibbles for that reason.

Is fd00::/8 like 10.0.0.0/8?

Partly. Unique local addresses (fc00::/7, in practice fd00::/8) are for use inside a site and are not routed on the Internet, like RFC 1918 ranges. Unlike 10.0.0.0/8, each site picks a random 40-bit Global ID (RFC 4193), so two networks that merge later are unlikely to clash. Neither is a firewall: being unrouted on the Internet is all a ULA guarantees.

Missing something?Suggest a feature or a new tool