About SpeedMeter
SpeedMeter is a free internet speed test built on one idea: tell people exactly what is being measured, and how, instead of showing a number and leaving them to guess.
Why another speed test?
Most speed tests are opaque. They give you a figure with no indication of how many connections were opened, whether the ramp-up was counted, or what the number actually represents. When the result disagrees with what you are paying for, there is no way to tell whether the problem is your Wi-Fi, your provider, or the test itself.
We publish the method. Every choice made in the measurement is described on this page and in our guides, so you can judge the result for yourself.
How the measurement works
Ping and jitter
Twelve very small requests are timed in sequence. The fastest round trip becomes your ping — the fastest is used rather than the average because it represents the path with the least queuing. Jitter is the mean difference between consecutive round trips, which is what makes calls stutter and games feel unpredictable.
Download
Several parallel connections pull randomly generated data for ten seconds. Parallel connections matter: a single stream is limited by round-trip time and TCP window size, and on a long-distance link one stream can show a fraction of the real capacity. Real workloads — a browser loading a page, a video player buffering — use many connections, so several is the honest way to measure.
The data is random because random data cannot be compressed. If we sent zeros, any compression along the path would give a wildly inflated result.
Upload
Several parallel connections push random data back for ten seconds, sampling genuine upload progress as the bytes leave rather than just timing whole requests.
The exact number of connections, and how much each one asks for, depends on the backend. A server with no per-request limit is best served by a few large transfers; a serverless host caps each request at a few megabytes, so the test uses more connections and more of them. The client asks the server what it can handle and sizes the run accordingly.
Discarding the ramp-up
Connections do not start at full speed. TCP slow start means the first seconds of any transfer are slower than the steady state. We discard the first two seconds of the download stage and the first two and a half of the upload stage, then average what remains. Without that, every result would read low.
What we do not do
We do not store your results, require an account, or ask for any personal detail. Your last ten tests live in your own browser and nowhere else. The full detail is in our privacy policy.
How the site is funded
SpeedMeter is free to use and supported by advertising. Running a speed test is unusually expensive to host — every test moves hundreds of megabytes — and ads are what covers that.
We deliberately hold ads back while a test is running. An ad loading mid-measurement would compete for the same bandwidth we are trying to measure and would make your result read low. Accuracy comes first.
Contact
SpeedMeter is operated by SpeedMeter. For corrections, questions or feedback, see the contact page.