IPv6 Cheat Sheet
Every IPv6 prefix length from /0 to /128, with the number of addresses it holds, how many /64 subnets fit inside it, its netmask, and what it is conventionally used for. The table opens on the lengths networks are actually cut on — expand it for all 129.
| Prefix | Addresses | /64 subnets | Netmask | Typical use |
|---|---|---|---|---|
| /128 | 1 | — | ffff:ffff:ffff:ffff:ffff:ffff:ffff:ffff | A single address — a host route or a loopback |
| /127 | 2 | — | ffff:ffff:ffff:ffff:ffff:ffff:ffff:fffe | Point-to-point link between two routers (RFC 6164) |
| /126 | 4 | — | ffff:ffff:ffff:ffff:ffff:ffff:ffff:fffc | Four addresses — the IPv4 /30 habit carried over to a router link |
| /124 | 16 | — | ffff:ffff:ffff:ffff:ffff:ffff:ffff:fff0 | 16 addresses — small static assignment |
| /120 | 256 | — | ffff:ffff:ffff:ffff:ffff:ffff:ffff:ff00 | 256 addresses — the IPv6 size of an IPv4 /24, used for static server ranges |
| /112 | 65,536 | — | ffff:ffff:ffff:ffff:ffff:ffff:ffff:0 | Sometimes used for numbered infrastructure links |
| /96 | 4,294,967,296 | — | ffff:ffff:ffff:ffff:ffff:ffff:: | Translation and mapping prefixes (::ffff:0:0/96, 64:ff9b::/96) sit here |
| /80 | 281,474,976,710,656 | — | ffff:ffff:ffff:ffff:ffff:: | Below the LAN boundary — breaks SLAAC, occasionally seen in labs |
| /64 | 1.84 × 10¹⁹ | 1 | ffff:ffff:ffff:ffff:: | One LAN. The standard subnet size, and the smallest that SLAAC works on |
| /60 | 2.95 × 10²⁰ | 16 | ffff:ffff:ffff:fff0:: | Smaller residential assignment — 16 /64 LANs |
| /56 | 4.72 × 10²¹ | 256 | ffff:ffff:ffff:ff00:: | The usual residential or small-business assignment — 256 /64 LANs |
| /52 | 7.55 × 10²² | 4,096 | ffff:ffff:ffff:f000:: | Nibble-aligned split of a site /48, tidy for reverse DNS |
| /48 | 1.2 × 10²⁴ | 65,536 | ffff:ffff:ffff:: | The standard site allocation — one organisation or campus, 65,536 /64s |
| /44 | 1.93 × 10²⁵ | 1,048,576 | ffff:ffff:fff0:: | Large enterprise or a POP inside an ISP |
| /40 | 3.09 × 10²⁶ | 16,777,216 | ffff:ffff:ff00:: | Aggregation boundary inside a large ISP; a nibble above /44 |
| /36 | 4.95 × 10²⁷ | 268,435,456 | ffff:ffff:f000:: | Regional slice of an ISP allocation |
| /32 | 7.92 × 10²⁸ | 4,294,967,296 | ffff:ffff:: | The standard ISP allocation — enough for 65,536 customer /48s |
| /29 | 6.33 × 10²⁹ | 34,359,738,368 | ffff:fff8:: | Common RIR allocation to an ISP, sized to allow later growth |
| /28 | 1.26 × 10³⁰ | 68,719,476,736 | ffff:fff0:: | Large ISP allocation from an RIR |
| /24 | 2.02 × 10³¹ | 1,099,511,627,776 | ffff:ff00:: | Very large ISP or national network allocation |
| /20 | 3.24 × 10³² | 17,592,186,044,416 | ffff:f000:: | Sub-allocation inside an RIR |
| /16 | 5.19 × 10³³ | 281,474,976,710,656 | ffff:: | Very large block; 2002::/16 (6to4) and ff00::/8's neighbours live at this scale |
| /12 | 8.3 × 10³⁴ | 4.5 × 10¹⁵ | fff0:: | An RIR's allocation from IANA |
| /10 | 3.32 × 10³⁵ | 1.8 × 10¹⁶ | ffc0:: | fe80::/10, link-local — every interface has one |
| /7 | 2.65 × 10³⁶ | 1.44 × 10¹⁷ | fe00:: | fc00::/7, the unique local (ULA) block |
| /3 | 4.25 × 10³⁷ | 2.3 × 10¹⁸ | e000:: | 2000::/3, the global unicast block the registries allocate from |
| /0 | 3.4 × 10³⁸ | 1.84 × 10¹⁹ | :: | Default route — matches every IPv6 address |
IPv6 has no broadcast address and no reserved network address, so every address in a prefix is usable. On a /64 the first address is also the subnet-router anycast address (RFC 4291).
Why IPv6 subnetting is different
IPv4 subnetting is an exercise in conservation: you work out the smallest prefix that fits the hosts you have, because addresses are scarce. IPv6 removes that constraint entirely, and with it the reason to size subnets to their hosts. Every LAN gets a /64 — a segment with two devices on it and one with two thousand are the same size — and the question becomes how many /64s you have to hand out rather than how many addresses.
The second difference is that IPv6 prefixes are cut on nibbles: four-bit steps, so /48, /52, /56, /60, /64. A nibble is one hex digit, and reverse DNS under ip6.arpa delegates one digit at a time — so a prefix that is a multiple of four gets its own reverse zone, and one that is not has to be served out of the enclosing zone. That is why you will see /56 and /48 everywhere and /50 essentially nowhere.
The prefix lengths that matter
- /32 — The standard allocation an RIR makes to an ISP — 65,536 customer sites at a /48 each.
- /48 — One site or organisation. 65,536 /64 LANs, and its own reverse DNS zone.
- /56 — The usual home or small-business delegation. 256 /64 LANs, which is more than a house needs and exactly the point.
- /64 — One LAN, and the smallest prefix SLAAC works on. Use it for every segment.
- /127 — A router-to-router link (RFC 6164), where there is nothing to autoconfigure.
Private and reserved IPv6 space
IPv6 has no RFC 1918 and no NAT to go with it. What it has instead:
- fc00::/7 — unique local addresses (RFC 4193), the closest equivalent to private space. Routable inside your network, dropped at its edge.
- fe80::/10 — link-local. Every IPv6 interface has one whether you configured it or not, and neighbour discovery runs over it.
- 2000::/3 — global unicast: the public internet, and the only space a registry allocates from.
- 2001:db8::/32 and 3fff::/20 — documentation (RFC 3849 and RFC 9637). Safe to put in a diagram, never routed.
- ff00::/8 — multicast, which in IPv6 does the work broadcast used to.
Work through a prefix
The IPv6 subnet calculator takes any address and prefix length and shows the boundaries, the /64 count, the reverse DNS zone, and — for public space — which network announces it in BGP. The IPv4 CIDR cheat sheet covers /0 to /32 on the other side.