Hosting speed,
seen in the wild.
Compare observed response times across delivery providers, grouped by resolver region, with a minimum sample before timings appear.
Check your site’s responseObserved delivery times. No claim of a worldwide winner.
Shared context. Clearer comparisons.
November 2026 Updated Nov 2, 2026, 8:19 AM UTC
Not enough measurements to compare providers yet.
We need at least 10 different websites per provider in the same measured region before showing response times. You can check an individual website now.
Check a website ↗Browse companies ↗
How these comparisons work
Useful context.
Honest limits.
These are anonymous observed delivery response times from newly completed eligible full Bare Audits. They include the network and CDN in the response path. They are not isolated origin-host timings or a scientific ranking of worldwide hosting speed.
What qualifies
One fresh, full mobile report must have HTTP 200 HTML for its homepage and all three Lighthouse scores averaging at least 75. Cached report views and the standalone speed checker do not add samples.
How samples count
We require 10 distinct normalized hostnames per provider and resolver region. Each hostname contributes once globally per month. Private keyed fingerprints deduplicate hostnames (www and apex count as one); names, paths and report links stay hidden. Collection starts without historical backfill.
Read the ranges
Times are grouped into 100 ms ranges, including the lower bound and excluding the upper bound. The median and 75th percentile are bucket estimates; the final bucket means 30,000 ms or more. Small differences may disappear within a bucket.
Keep the context
We group scans by the Cloudflare DNS region observed in both safety lookups for the final URL. This resolver location does not prove where the scanner or origin ran. Cache state, application behavior and visitor mix vary; selection bias is possible. Provider roles are inferred and the origin may remain unidentified.
A slower response can delay the browser’s start and affect FCP or LCP indirectly. Those metrics depend on many factors. These ranges cannot tell you an exact Lighthouse point loss or prove that a particular host is bad.
Further readingOptimize Time to First Byte ↗How Lighthouse performance scoring works ↗