01 — Network tool
Internet speed test
Measure your download and upload speed, ping and jitter in about 25 seconds. No app, and nothing is stored.
Ready
—Mbps
- 01Ping
- 02Download
- 03Upload
- Server
- Ashburn, US-EastResults are lower the farther you are from it.
- Data usage
- About 100–250 MB per testMind metered mobile plans.
- Duration
- ~25 seconds: 10 s download, 10 s upload
Your results
- Download
- —Mbps
- Upload
- —Mbps
- Ping
- —ms
- Jitter
- —ms
What this result is good for
What it means
Download is how fast data reaches you: streaming, web pages, game updates.Upload is how fast you can send: video calls, cloud backups, posting videos. Both are measured in megabits per second (Mbps). Eight megabits make one megabyte, so 100 Mbps moves about 12.5 MB every second.
Ping (latency) is the time a tiny request needs to reach the test server and come back, in milliseconds. It decides how responsive things feel: a shot landing in a game, the other person reacting on a call, a page starting to load. Jitter is how much ping wobbles from one moment to the next. For gaming and calls, a low and steady ping matters more than raw speed.
Rules of thumb
- Under 10 Mbps: browsing, email and music are fine, one HD stream works; several people at once will notice.
- 25–100 Mbps: 4K streaming on a screen or two, HD video calls, smooth gaming if ping is low.
- 100–500 Mbps: a busy household with many devices, large game downloads in minutes.
- 500 Mbps and up: gigabit class. At this point Wi-Fi or the device itself is usually the limit, not the line.
- Ping: under 20 ms is excellent, under 50 ms is good for gaming, under 100 ms is fine for calls. Jitter under 10 ms is steady.
Why your result may be lower than your plan
Providers advertise the maximum speed delivered to your router. Wi-Fi is the most common bottleneck: walls, distance and crowded channels can easily halve throughput. Devices updating in the background, VPNs, old network adapters and busy evening hours all pull the number down. For the most honest reading, test on a cable with nothing else running, then repeat on Wi-Fi in the rooms you actually use.
Distance matters too. Our test server is in Ashburn, US-East. If you are far away, ping rises and each connection ramps up more slowly, so a nearby server will usually show better numbers.
How it works
The test runs entirely in your browser, in a Web Worker so the page stays responsive. The methodology is inspired by the open-source LibreSpeedproject:
- Ping and jitter. A warm-up request opens the connection first, so DNS lookup and the TLS handshake are not counted. Then 20 tiny requests are timed one after another with the browser's Resource Timing API. Ping is the median, so a single slow outlier can't skew it. Jitter is the average difference between consecutive samples.
- Download. Six parallel connections stream random data for about 10 seconds. The first 1.5 seconds are discarded, because TCP starts slowly and needs time to ramp up; the rest is averaged. Request sizes adapt to your speed, so fast links don't flood the server with tiny requests and slow links don't fetch data they can't finish.
- Upload. Five parallel connections send blocks of random bytes, generated with the Web Crypto API so nothing along the way can compress them, using the same time window and warm-up rule.
The figures are payload throughput: the useful data that arrived, without TCP/IP header overhead, which is why they land a few percent below the raw line rate. Nothing is stored. The numbers live in this tab until you copy them with Copy result.