Explainer · networking

IPv4 vs IPv6: the differences that actually matter

Everyone knows IPv6 has “more addresses.” Almost nobody can say what actually changes when a connection uses it. This is the practical version: the format differences, the notation rules that trip people up, why NAT mostly goes away, and the short list of things you genuinely need to care about.

Why IPv6 exists at all

IPv4 uses a 32-bit address space. Two to the power of thirty-two is roughly 4.3 billion, and that felt limitless when the protocol was standardised in 1981 — there were a few hundred computers on the network. It stopped feeling limitless once every phone, laptop, thermostat and doorbell wanted its own address. The five Regional Internet Registries handed out their final large IPv4 blocks between 2011 and 2019, and today a fresh IPv4 address has to be bought on a secondary market for tens of dollars apiece.

IPv6 was designed in the 1990s specifically to end that scarcity. It moves from 32 bits to 128 bits per address. That is not four times bigger; it is two to the power of ninety-six times bigger — around 340 undecillion addresses, a number with 39 digits. The whole point is that address exhaustion simply stops being a design constraint. You can hand a globally-unique address to every device and never think about running out again.

The address format, side by side

An IPv4 address is written as four decimal numbers, each 0–255, separated by dots — “dotted decimal.” Each number is one byte (8 bits), so four of them make up the 32-bit total. A familiar example:

192.0.2.146

An IPv6 address is written as eight groups of four hexadecimal digits, separated by colons. Each group is 16 bits, so eight groups give the 128-bit total. Hexadecimal is used because writing 128 bits in decimal would be unwieldy. A full, uncompressed example:

2001:0db8:0000:0000:0000:ff00:0042:8329

That is a lot to type, which is exactly why IPv6 has two compression rules that let you write the same address far more compactly.

Notation rules that trip people up

Leading-zero suppression

Within any group, leading zeros can be dropped. So 0db8 becomes db8, 0000 becomes 0, and 0042 becomes 42. Applying that to the address above gives:

2001:db8:0:0:0:ff00:42:8329

The double colon

A single run of consecutive all-zero groups can be replaced with a double colon, ::. The parser fills the gap back out with as many zero groups as needed to reach eight total. So the address collapses to:

2001:db8::ff00:42:8329

The critical rule: :: may appear only once in an address. If it appeared twice, there would be no way to know how many zero groups belong to each gap. A couple of well-known shorthands fall straight out of this: the IPv6 loopback address (equivalent to IPv4's 127.0.0.1) is 0:0:0:0:0:0:0:1, which compresses all the way down to ::1, and the “unspecified” all-zeros address is simply ::.

Prefixes and the /64 subnet

IPv6 uses CIDR-style prefix notation just like IPv4: a slash followed by the number of leading bits that identify the network. The convention that surprises people is that a single LAN is almost always a /64 — that leaves 64 bits, or about 18 quintillion addresses, for hosts on one network segment. It feels absurdly wasteful coming from IPv4, where you might carve a home network out of a /24 with 254 usable hosts. In IPv6 the abundance is deliberate: fixing the host portion at 64 bits is what makes stateless address autoconfiguration work. If you want to sanity check prefix maths in either protocol, the CIDR calculator handles both.

NAT vs end-to-end addressing

This is the difference that changes how the internet actually behaves. Because IPv4 addresses are scarce, almost every home and office hides many devices behind a single public address using Network Address Translation. Your devices get private addresses (the 192.168.x.x and 10.x.x.x ranges), and the router rewrites the source address on the way out and reverses it on the way back. Many ISPs now add a second layer of NAT on top of that, sharing one public address between hundreds of customers — the topic of our carrier-grade NAT explainer.

IPv6 restores the internet's original design, where every device has its own globally-routable address and there is no translation in the middle. This is called end-to-end addressing. It makes inbound connections possible again — hosting a service at home, direct peer-to-peer links, and protocols that dislike NAT all become simpler. It does not mean every device is wide open: home routers ship a stateful IPv6 firewall that drops unsolicited inbound traffic by default, giving you the same practical protection NAT provided as a side effect, but now as an explicit, intended security control rather than an accident of address scarcity.

Dual-stack, and why IPv4 refuses to die

IPv4 and IPv6 are not interoperable at the packet level: an IPv6-only host cannot talk directly to an IPv4-only host. There is no gradual conversion where addresses slowly turn into IPv6. Instead, the migration strategy is dual-stack: most connections run both protocols simultaneously, and the operating system chooses which to use for each destination. Your laptop might reach one site over IPv6 and the next over IPv4 without you ever noticing.

That is why IPv4 persists decades after IPv6 shipped. As long as any meaningful number of servers, CDNs, or corporate networks remain IPv4-only, every client has to keep speaking IPv4 to reach them — so nobody can switch off the old protocol, so there is no forcing event. The result is a long, stable coexistence rather than a clean cutover. Adoption is nonetheless climbing steadily; if you want to see where it stands, our IPv6 adoption status page tracks the trend rather than repeating the explainer you are reading now.

Performance and Happy Eyeballs

A common myth is that IPv6 is faster. A bit on the wire does not care which protocol labelled it. Where IPv6 can genuinely win is the path: an IPv6 connection often skips the carrier-grade NAT translation that an IPv4 connection has to pass through, and fewer boxes in the middle can mean marginally lower latency. But this is situational, not a law.

To avoid users ever waiting on a slow or broken protocol, browsers and operating systems implement an algorithm called Happy Eyeballs (RFC 8305). When a hostname resolves to both an IPv6 and an IPv4 address, the client starts connecting over IPv6 but, after a very short head start, races an IPv4 attempt in parallel and uses whichever connection completes first. If IPv6 is misconfigured or slow, the fallback is nearly instant and invisible. This is why the transition has been so smooth for ordinary users: the software quietly hedges its bets on every single connection.

Privacy considerations

IPv6's end-to-end model raised a reasonable worry: if every device has its own permanent global address, could websites track a specific laptop or phone across networks the way a cookie does? The original autoconfiguration method, SLAAC (stateless address autoconfiguration), made this worse in early designs by deriving the host portion of the address from the network card's MAC address, producing a stable identifier that followed the device around.

The fix is privacy extensions (RFC 8981), which every major operating system now enables by default. Instead of a MAC-derived address, the device generates a random host portion for outbound connections and rotates it regularly — often daily. Your device keeps a stable address for inbound services if it needs one, but the address the outside world sees on your outbound traffic changes over time, defeating naive device-level tracking. The practical upshot is that a modern IPv6 host is no more trackable by its address than an IPv4 host behind NAT. You can see the actual addresses your connection presents on the what is my IP page.

Do I actually need to care?

For most people, honestly, no — dual-stack and Happy Eyeballs mean everything just works, and the operating system makes the right choice for you invisibly. But there are a few situations where knowing the difference pays off:

Everyone else can file IPv6 under “good to understand, safe to ignore in daily use.” The internet has spent decades making the difference invisible on purpose.

Frequently asked questions

Is IPv6 faster than IPv4?

Not inherently — a packet does not move faster for being IPv6. What can make it feel faster is the route: IPv6 traffic often bypasses the extra carrier-grade NAT layer that many IPv4 connections pass through, trimming a hop. Browsers also race both protocols with Happy Eyeballs and use whichever answers first, so you transparently get the quicker path. On a healthy network the difference is usually negligible.

Do I need to disable IPv4 once I have IPv6?

No, and you should not. Nearly everything runs dual-stack, with both protocols active and the operating system picking the best one per destination. Disabling IPv4 would break access to the large share of the internet that is still IPv4-only. Dual-stack is the intended steady state, not a temporary workaround.

Does using IPv6 expose my devices to the internet?

A globally-routable address is not the same as a reachable one. Home routers ship with a stateful IPv6 firewall that drops unsolicited inbound connections by default, just as IPv4 NAT did — nothing is open unless you deliberately open it. And for outbound traffic, privacy extensions rotate your address so you are not trackable by a fixed device identifier.

Related reading