Reference

Public DNS resolvers compared

What each public resolver filters, the addresses it publishes, and when we last checked those details against the provider’s own documentation. There is no “best” row in this table, because which resolver performs best depends on where you are and how your network routes — that is what the benchmark is for.

How to read this table

Every value below is taken from the operator’s published documentation, not from third-party lists, and every row carries the date we last checked it. Rows marked Not re-verified are carried over from an earlier release of this site and have not been confirmed against a primary source in the current pass — they are shown for completeness and are deliberately excluded from the benchmark until someone re-checks them. 18 of the entries are currently in that state.

The table is generated directly from RESOLVER-REGISTRY.json, the same file the benchmark reads. It is not possible for this page to describe a resolver differently from the way the tool tests it.

The comparison

Choosing between them

Filtering is a product decision, not a quality ranking

Most operators publish several variants of the same service. A “family” endpoint is not a better version of the unfiltered one; it is a different product that returns different answers. Picking a filtering resolver means accepting that some lookups will be answered with a block page or an NXDOMAIN, including occasional false positives. Picking an unfiltered one means accepting that nothing will be blocked on your behalf.

“No logs” is a claim, not a measurement

Several operators here state that they do not retain query logs. Nothing on this page verifies that claim — it cannot be verified from outside. What can be said with confidence is the structural fact: whichever resolver you configure will receive every domain name your device looks up. Read the linked privacy policy and decide whose stated policy and jurisdiction you prefer.

ECS changes the privacy trade-off

Some variants send EDNS Client Subnet, which forwards a truncated version of your network address to authoritative name servers so that content delivery networks can return a geographically closer answer. That often produces faster downloads and a less private lookup. Neither choice is wrong; they optimise for different things.

Speed is local

A resolver’s response time depends on which of its sites your traffic actually reaches and how your provider routes there. A resolver that is quickest on one connection can be middling on another in the same city. That is why this page does not rank by speed, and why the benchmark measures from your browser instead of quoting a table.

Changes in this release

• dns0.eu removed. The operator states the service has been discontinued and points users to DNS4EU or NextDNS. Three dns0.eu entries were being offered by this site as working resolvers; they have been retired.
• DNS4EU added. Four EU-operated profiles, verified against the operator’s documentation, now cover the European option that dns0.eu previously filled.
• Mullvad’s secondary address corrected. The previous release listed an IPv4 address for Mullvad that does not appear in Mullvad’s published table. It has been removed rather than guessed at.
• CleanBrowsing endpoint paths corrected to the documented form, which includes a trailing slash.

Related

• Run the benchmark against your own connection
• What DNS-over-HTTPS actually changes
• How this benchmark measures, and what it cannot tell you

Last reviewed 2026-08-15. Provider configuration comes from RESOLVER-REGISTRY.json, which records the official source and verification date for every entry.