IPv4 vs IPv6 proxies: what is the difference?

IPv4 and IPv6 proxies at a glance

Why IPv4 addresses are scarce

IPv4 has about 4.3 billion addresses, far fewer than devices online. The regional registries have handed out their free pools: the RIPE NCC announced on 25 November 2019 that it had made its final /22 allocation, and since then it has distributed only small amounts of returned space through a waiting list, one /24 per eligible new member.

Networks cope in two ways. They buy and lease addresses on a secondary market, which puts a price on every IPv4 address a proxy provider holds. And internet providers share one public address among many subscribers through carrier-grade NAT, which is common on mobile networks.

Will the target accept IPv6?

An IPv6-only exit can only connect to a host that has an IPv6 address. Check with dig AAAA example.com +short: no answer means the site is reachable over IPv4 only, and an IPv6 proxy cannot reach it unless the provider translates to IPv4 — in which case the target sees an IPv4 address anyway.

With an HTTP proxy, the proxy resolves the host name and chooses the address family for the outbound connection, not your client. On a dual-stack proxy, the same request can leave over IPv4 or IPv6 depending on the provider's configuration, so ask which one is used and test it. Google publishes per-country statistics on the share of its users reaching it over IPv6; those describe client adoption, not whether any given website supports IPv6.

Why a /64 counts as one client

RFC 4291 requires 64-bit interface identifiers for nearly all IPv6 unicast addresses, so a /64 is the standard size of a single network segment. Devices in it can pick any of its 2⁶⁴ addresses, and change them often for privacy. RFC 6177 recommends that end sites receive more than one /64 — it names a /56 as a reasonable default for home users — and no longer recommends handing out single /128 addresses.

For a site operator, that means one IPv6 address says almost nothing about who is connecting: the same household or server can present thousands. Rate limiting per address would be trivial to sidestep, so limits are applied per prefix — a /64, sometimes a /56 or /48. Traefik documents this directly: its ipv6Subnet setting transforms an address into the first address of its subnet before counting.

The practical consequence for a buyer: a pool of a million IPv6 addresses from a handful of prefixes in one ASN behaves, for rate limiting, like a handful of clients. Spread across subnets and networks is what matters, not the raw count.

When an IPv6 proxy makes sense

• The targets publish AAAA records and you have confirmed they serve the same content over IPv6.
• The workload is high-volume and the endpoints are tolerant — public APIs, your own infrastructure, CDNs.
• You are testing your own IPv6 deployment from outside your network.

Avoid it when targets are IPv4-only, when they limit strictly per prefix, or when what the target shows depends on the visitor being on a consumer network. If you hit limits, the remedy is slower, better-spread traffic — see handling 429 rate limits.

What ARC Proxies offers

We do not sell a separate IPv6 proxy product. Our plans are defined by network type — residential, rotating and static ISP and mobile — not by IP version. If your project specifically needs IPv6 exits, ask before you buy rather than assuming.