📶 Internet Speed Test — Free, In-Browser, No Sign-Up

About

The number this test reports is throughput between your device and a Cloudflare edge server, and that is deliberately not the same thing as the line rate your ISP advertises. A 500 Mbps plan describes the capacity of the link into your home; what actually reaches a browser tab is that capacity minus protocol overhead, minus whatever your Wi-Fi is losing, minus bandwidth other devices in the house are already using, minus any throttling your browser applies to a background tab. Seeing 380 Mbps on a 500 Mbps plan over Wi-Fi is ordinary, not a fault.

Two design choices make the measurement more honest than a naive one. First, the test opens four parallel connections instead of one. A single TCP stream cannot fill a high-bandwidth, high-latency link — the congestion window simply does not grow fast enough — so single-stream tests systematically under-report fast connections. Second, the first 800 milliseconds of transfer are discarded. TCP begins conservatively and ramps up, so bytes counted during that ramp would drag the average down and make every connection look slower than it is. What you see is the throughput after the connection has reached full speed.

Ping is reported as the minimum of fifteen round trips rather than the average. Queuing delay, garbage collection and scheduler noise can only ever make a sample slower, never faster, so the smallest sample is the one closest to the true round-trip time. Jitter is the mean absolute difference between consecutive samples. For real-time applications jitter matters more than raw speed: a 100 Mbps connection with 40 ms of jitter will produce worse video calls than a 50 Mbps connection with 3 ms of jitter, because the receiver has to buffer against the variation.

One honest limitation deserves stating plainly. If you are measuring in order to file a complaint with your ISP about not receiving contracted speeds, this test is not the instrument to use. In Korea the official measurement for that purpose is the NIA internet quality test, which runs against government-designated servers under controlled conditions and produces a result carriers are obliged to recognise. This test is for the everyday question — is my connection working normally right now, and is the problem my line or the site I am trying to use — not for a contractual dispute.

Results also move around more than people expect. Wi-Fi signal strength, someone else streaming in the next room, VPN encryption overhead, and even a busy CPU can each move the figure by tens of percent. A single reading is a snapshot. Running the test three times and comparing, ideally once over Wi-Fi and once over a wired connection, tells you far more than any one number.

How to use

1. Close other tabs and pause any downloads, streaming or cloud backups. Anything else using the connection is competing for the same bandwidth and will show up as a lower result.
2. Press Start. Latency is measured first with fifteen small round trips, then download throughput for eight seconds, then upload for six seconds.
3. Read download and upload in Mbps, ping and jitter in milliseconds. The test server's Cloudflare location code is shown so you can tell how far the traffic travelled.
4. Run it two or three times. If the numbers swing widely between runs, the instability itself is the finding — that usually points at Wi-Fi interference rather than the line.
5. To separate Wi-Fi problems from line problems, repeat the test with an Ethernet cable. If wired is fast and wireless is not, the router or the wireless environment is the bottleneck.

FAQ

Why is my result lower than the speed my ISP advertises?

The advertised figure is the capacity of the line into your home, measured under ideal conditions. What reaches a browser tab is that figure minus protocol overhead, Wi-Fi losses, bandwidth used by other devices, and any browser throttling. A result of 70 to 85 percent of the advertised rate over Wi-Fi is normal. If you consistently see far less than half, test again over a wired connection before concluding the line is at fault.

Can I use this result to complain to my ISP?

No, and we would rather say so than let you find out later. Carriers accept measurements taken under controlled conditions against designated servers. In Korea that is the NIA internet quality measurement service. This test measures throughput to the nearest Cloudflare edge, which is excellent for answering "is my connection behaving normally right now" but is not a contractual instrument.

What are good ping and jitter values?

For ping, under 20 ms is excellent, 20 to 50 ms is fine for almost everything including most gaming, and over 100 ms becomes noticeable in voice calls and competitive games. For jitter, under 5 ms is good and under 20 ms is usually tolerable; above 30 ms you will hear it as choppy audio in calls even if your download speed looks high. Jitter is the value most people never check and the one that most often explains a connection that "feels" bad while testing fast.

Why is my upload so much slower than my download?

On most consumer connections that is by design. Cable and DSL are asymmetric — capacity is allocated mainly to download because that is what typical households consume — so an upload figure that is a tenth of the download figure is normal on those technologies. Fibre plans are often symmetric, so on fibre a large gap is worth investigating. If you work from home and upload matters, the ratio is a more useful thing to check than the headline download number.

Does this send my files or personal data anywhere?

No files and no personal data. Unlike the rest of Toolio, this tool genuinely does use the network — that is what it measures — but the bytes it exchanges are randomly generated for the test and discarded on arrival. Nothing from your device is read or uploaded. The test does need an internet connection to work, so unlike our offline tools it will not run with the network disabled.

Why do I get a different number every time I run it?

Because network conditions genuinely change second to second. Wi-Fi signal fluctuates, other devices start and stop using bandwidth, and the path to the test server can be more or less congested. Variation of 10 to 20 percent between runs is expected. Swings much larger than that are themselves diagnostic and usually point at wireless interference — try the same test over Ethernet and see whether the spread disappears.

Is there a limit on how many tests I can run?

There is a cap of 40 tests per day per network address, plus a site-wide daily capacity limit. Each run transfers a meaningful amount of data, and Toolio's server budget is shared across every tool that needs an API — so an unlimited speed test would starve the rest of the site. Normal use, checking a connection a few times a day or comparing Wi-Fi against wired, will never come close to either cap. If you see a capacity message, the site has hit its daily budget and the test will work again tomorrow.