IPv4 vs IPv6 Proxies: What Is the Difference?
IPv6 gives you addresses by the thousand at a low price — but only where the target speaks IPv6. Here is how to decide.
IPv4 and IPv6 are two generations of the internet’s addressing system. IPv4 addresses are 32 bits long, giving about 4.3 billion in total — far fewer than the number of connected devices. IPv6 uses 128 bits, an astronomically larger space.
What changes for proxies
- Price: IPv4 addresses are scarce and priced individually; IPv6 addresses are abundant, so packs are cheap.
- Reach: an IPv4 proxy reaches essentially everything; an IPv6-only exit reaches only IPv6-capable destinations.
- Rate limits: some networks limit a whole /64 subnet as if it were a single user, which can reduce the benefit of many addresses.
- Detection: heavily defended sites may treat IPv6 ranges differently from IPv4 ones.
How to check if a site supports IPv6
Look for an AAAA record. The DNS lookup tool on this site lets you query it: if the domain returns AAAA records, IPv6 clients can reach it; if it only returns A records, an IPv6-only proxy cannot.
Dual stack and fallback
Many clients are dual-stack and try IPv6 first, falling back to IPv4. Through an IPv6 proxy the fallback is not available, so unsupported sites simply fail.
Which should you buy?
Buy IPv6 when your targets are IPv6-ready and you need large numbers of addresses at low cost. Buy IPv4 — datacenter, ISP, residential or mobile — when you need universal reach or a particular reputation.
Proxy toolsDNS LookupLook up DNS records through public resolvers. Useful for checking IPv6 support, mail setup and domain ownership records.ExploreWhat a /64 really contains
An IPv6 address is 128 bits, written as eight groups of four hexadecimal digits. By convention the first 64 bits identify the network, called the prefix, and the last 64 bits identify an interface inside it, so a single /64 holds 2^64, or about 1.8 × 10^19, addresses. That is roughly 4.3 billion times the whole IPv4 internet, which has 2^32 addresses in total. The IPv6 plans on this site deliver 100, 500, 2,500 or 15,000 addresses from one /64 subnet. Even the largest pack is a vanishing slice of its own prefix, and every address you receive shares the same first 64 bits.
A few writing rules help when you read a delivered list. Leading zeros in a group may be dropped, and one run of all-zero groups may be replaced by a double colon, so 2001:0db8:0000:0000:0000:0000:0000:0001 becomes 2001:db8::1. The 2001:db8::/32 range is reserved for documentation, which is why the examples here use it. In a URL an IPv6 literal goes in square brackets, because the colons would otherwise be confused with the port separator.
curl -x http://user:pass@[2001:db8::1]:8080 -I https://example.comWhy sites defend a prefix, not an address
Operating systems commonly generate temporary IPv6 addresses (privacy extensions), so one household device may use many addresses over a day, all inside the same prefix. A site that blocked or rate limited single addresses would be chasing noise, so defenders aggregate: limits and reputation are often applied per /64, and some systems use wider prefixes such as /56 or /48.
The arithmetic shows the effect. Suppose a target allows 60 requests per minute per client (illustrative) and counts clients per address. A hundred IPv4 addresses then give you 6,000 requests per minute. If it counts per /64, all 15,000 addresses of an IPv6 pack together still get 60 per minute, and the pack size changes nothing. Most sites do not publish which way they count, so the practical answer is to test a small pack, as described below.
Four checks before you buy
- Query the AAAA record of the main hostname with the DNS lookup tool. No AAAA record means an IPv6-only exit cannot reach the site at all.
- Repeat the query for every hostname the page uses: the API subdomain, the image and script CDNs, the login host. One missing AAAA record is enough to break part of a page.
- Look for IPv4 literals. URLs such as http://203.0.113.5/, or configuration files that hand out IPv4 addresses, cannot be reached from an IPv6-only exit.
- Run a small test through the proxy at your intended rate and compare the success rate with an IPv4 baseline.
You can automate the first two checks. The helper below tests whether a hostname has AAAA records. DNS answers for public sites are the same for the proxy as for you, so the result applies to both, but it checks resolution only, not whether a connection would succeed.
import socket
def has_ipv6(host):
try:
socket.getaddrinfo(host, 443, socket.AF_INET6, socket.SOCK_STREAM)
return True
except socket.gaierror:
return False
for h in ("example.com", "api.example.com"):
print(h, has_ipv6(h))Reading failures through an IPv6-only exit
- An immediate error such as bad gateway, host unreachable or no route: the destination has no AAAA record, or there is no IPv6 path from the proxy’s network. The exact wording depends on the proxy software.
- The main page loads but images or scripts fail: some of the page’s hostnames are IPv4-only.
- Success at first, then 403 or 429 across all addresses at once: the target limits or blocks at the prefix, not per address. Slow down, or accept that this target needs IPv4.
- The site shows the wrong language or region: IPv6 geolocation data can be less consistent than IPv4, and the location of these plans is not stated, so do not choose them where location matters.
Sizing an IPv6 pack
Because the prefix is often the limiting unit, size a pack by per-address limits only after you have established that the target counts per address. A worked example (illustrative): you need 3,000 requests per minute against a target that, in testing, allows 30 requests per minute per address. Dividing 3,000 by 30 gives 100 addresses, which is exactly the smallest pack and leaves no headroom. Either lower your target rate to leave a margin, or step up to 500 addresses, which serve the same load at 6 requests per minute each. Starting with the 100-address pack is the cheapest way to learn how the target behaves, and the pricing page lists current prices.
Remember the protocol constraint as well. IPv6 plans list HTTP and HTTPS, and SOCKS5 is not confirmed, so choose tools that accept an HTTP proxy URL.
Dual-stack clients, Happy Eyeballs and why they do not help here
Browsers and many libraries implement Happy Eyeballs (RFC 8305): they start an IPv6 attempt, a fraction of a second later an IPv4 attempt, and use whichever completes first. That logic lives in the client that resolves the name and connects. When you use a proxy, the proxy normally makes the connection to the destination, so the fallback decision belongs to the proxy, and an IPv6-only exit has no IPv4 path to fall back on. Your own client’s IPv4 and IPv6 behavior affects only the connection to the proxy itself. This is why the address you connect to and the address websites see are two separate questions, and why you should read the delivered details for both.
Frequently asked questions
Can NAT64 let an IPv6 proxy reach IPv4-only sites?
NAT64 with DNS64 is a technique some networks use to give IPv6-only clients access to IPv4 destinations. Whether the IPv6 exit used by these plans does this is not confirmed, so do not rely on it. Ask sales on WhatsApp before you build on an IPv4-only target.
How can I tell that my IPv6 addresses share one prefix?
Compare the first four groups of several addresses from your list. If they are identical, the addresses sit in the same /64, which is how these plans are delivered. A site that limits by prefix will treat them as a single client, so keep that in mind when you plan request rates.
Is an IPv6 proxy slower than an IPv4 one?
The protocol version itself does not make a connection slow. What matters is the route and the capacity behind the proxy. The IPv6 plans list 100 to 500 Mbit, but a listed figure is not a measurement, so test throughput from your own location against the targets you care about.