⚡

Key Latency Test

Measures how long you hold each key and how long you leave between presses, reported as median and 95th percentile. Read the caveat below before drawing conclusions about hardware.

What this isBrowser event timing — how long you hold keys and how long you leave between them. It is not hardware input lag, which no web page can measure.

Dwell time — how long you hold each key

Median

—

Average

—

95th percentile

—

Samples

Flight time — the gap between presses

Median

—

Fastest

—

95th percentile

—

Implied WPM

—

Start typing

Ordinary sentences, not random mashing — dwell time only means something on real text.

What a Browser Can and Cannot Measure

Start with the caveat, because most pages that call themselves a keyboard latency test get this wrong. A web page cannot measure your keyboard's hardware input lag. Real input lag is the time between your finger pressing the switch and the resulting change appearing on screen, and it accumulates across the switch actuation, the controller scan, the USB polling interval, the operating system input stack, the browser event loop, and finally the display. A browser only gets to see the last few of those, and it sees them through a timestamp the browser itself assigns. Measuring the whole chain needs a high-speed camera or a hardware probe.

What this page measures instead is real, useful, and clearly defined:

Dwell time — how long each key stays down, from keydown to keyup. This is a property of your typing, not your keyboard, and it is the single best indicator of typing efficiency.
Flight time — the gap between one key going down and the next. This is your rhythm, and the variability in it matters more than the average.
Percentiles — the median tells you what is typical; the 95th percentile tells you about your worst moments, which is where inconsistency actually lives.

Why percentiles rather than an average: an average hides the thing you want to see. Two typists with identical average dwell times can behave completely differently — one steady, one alternating between very short and very long holds. The gap between your median and your 95th percentile is the number that exposes that. A median of 90ms with a p95 of 120ms is a smooth typist. The same median with a p95 of 300ms means something is stalling regularly, usually a specific finger or a specific key reach.

Type at least thirty keystrokes of ordinary text before reading anything into the numbers, and preferably a couple of hundred. Percentiles are meaningless on small samples: with ten data points, the 95th percentile is just the second-slowest one you happened to produce.

One more measurement note. Browsers deliberately coarsen their high-resolution timestamps to make fingerprinting harder — Firefox rounds to whole milliseconds without cross-origin isolation, Chrome to about a tenth of one. At the tens-to-hundreds of milliseconds this page reports, that rounding is irrelevant. It would matter a great deal if we were claiming to measure hardware latency, which is part of why we are not.

If it is a fault rather than timing you are chasing: keys registering twice belong in the double typing test, keys that will not release belong in the stuck key test, and dropped inputs under load belong in the ghosting test.

Key Capabilities & Features

Dwell time per keystroke

Keydown to keyup for every key, which is the clearest single indicator of typing efficiency.

Flight time between presses

The gap from one key going down to the next — your rhythm, and a direct proxy for typing speed.

Median and 95th percentile

Percentiles rather than averages, because the gap between them is where inconsistency shows up.

Per-key breakdown

Ranks your slowest keys by median dwell, which usually points at one finger or one awkward reach.

Sample-size gate

Tells you when there is enough data for percentiles to mean anything, instead of reporting a p95 from ten samples.

Honest about its limits

States plainly that this is DOM event timing, not hardware input lag, and explains what does affect the latter.

How to Perform This Test

Click once on the page so the browser tab has keyboard focus.:
Type ordinary text for a minute or so — a paragraph you know by heart works best. Random mashing gives useless dwell times.:
Watch the sample counter. Percentiles need at least thirty samples before they mean anything, and a couple of hundred is better.:
Read your median dwell first. Between 60 and 100 milliseconds is efficient touch typing.:
Compare the median against the 95th percentile. A wide gap means occasional stalls rather than a general slowness.:
Check the per-key table for the keys with the highest median. Those are usually the reaches your fingers find awkward.:

Troubleshooting Common Hardware Issues

All values show a dash: no complete keydown-keyup pairs recorded yet. Press and release a few keys normally rather than holding them.
The 95th percentile looks wild: you probably have fewer than thirty samples. Keep typing — with ten samples the p95 is just the second-slowest one you happened to make.
Your dwell is very short and you are losing letters: you may be releasing before the switch fully registers. Try a slightly firmer press, or lower the actuation point if your board allows it.
One key has a much higher median than the rest: usually an awkward reach rather than a fault. Check the same key on the full keyboard test to rule out hardware.
Timings look identical to the millisecond: your browser is rounding timestamps for privacy. Harmless at this scale.
The numbers change if you switch to a different keyboard: expected, and interesting. Switch travel and actuation point genuinely change how long you hold keys.
You wanted hardware input lag: a browser cannot measure it, and any page claiming to is measuring something else. Hardware review sites publish instrumented end-to-end figures.

Reading Your Dwell and Flight Times

These ranges come from the keystroke-dynamics literature, where dwell and flight times have been studied for decades as a biometric. They are a guide to where you sit, not a target to chase.

Median dwell	Typical of	What it suggests
Under 60ms	Fast touch typists, 90+ WPM	Very light, efficient key release. Common on low-actuation switches.
60 – 100ms	Competent touch typing	The sweet spot. Keys are released as soon as they register.
100 – 150ms	Average typing	Slightly heavy, often from pressing past the actuation point.
150 – 250ms	Hunt-and-peck or careful typing	Keys held while the eyes find the next one.
Over 250ms	Deliberate or unfamiliar layout	Normal when learning a new layout or keyboard size.

Flight time maps almost directly to speed. Since a word averages five characters plus a space, your words-per-minute is roughly 60000 ÷ (median flight × 6). A 200ms median flight is about 50 WPM; 120ms is about 83 WPM; 100ms is about 100 WPM. If you want to check that against a proper measured score, the WPM test does it properly with real text and error accounting.

Shorter dwell is not automatically better. Very short dwell times on a keyboard with a high actuation point can mean you are releasing before the press fully registers, which shows up as dropped characters rather than speed. If your dwell is under 50ms and you are also losing letters, that is the likely explanation.

What Actually Reduces Keyboard Input Lag

Since this page cannot measure hardware lag, here is what genuinely affects it, roughly in order of how much difference each makes. This is the part people are usually looking for when they search for a latency test.

Polling rate. A 125Hz USB keyboard is sampled every 8ms; a 1000Hz one every 1ms. This is the largest single hardware contributor, and it is a published spec you can check before buying.
Wireless versus wired. Modern 2.4GHz dongles add roughly 1–2ms and are effectively indistinguishable from wired in use. Bluetooth adds considerably more and varies with interference. If latency matters, the dongle is fine and Bluetooth is not.
Actuation distance. A switch that actuates at 1.2mm registers sooner than one at 2.0mm, purely because your finger travels less. Adjustable Hall-effect and optical switches let you tune this.
Debounce time. Firmware waits a few milliseconds after contact to filter switch bounce. Lower debounce means lower latency and more risk of chatter — the trade-off is direct, and QMK boards let you set it.
Display and frame pacing. Often the biggest term of all, and nothing to do with the keyboard. A 60Hz display adds up to 16ms before your keystroke can possibly appear; a 240Hz display adds up to 4ms. V-sync and frame buffering add more.
Software in the path. Remapping tools, macro layers, remote desktop, virtual machines and KVM switches all insert themselves between the keyboard and the application. Any one of them can add more delay than every hardware factor combined.

For perspective: the difference between a good and a mediocre keyboard is a handful of milliseconds. The difference between a 60Hz and a 240Hz display is around 12. The difference made by running your input through remote desktop can be 50 or more. If you are optimising for responsiveness, work down the list in that order rather than starting with the keyboard.

If you want to compare your board against real measurements, hardware review sites publish end-to-end latency figures captured with instrumented equipment. Those numbers are trustworthy in a way no browser-based test can be, and that is worth saying plainly.

Frequently Asked Questions

Can a website measure my keyboard latency?

Not the hardware kind. Real input lag spans the switch, the controller, USB polling, the OS input stack, the browser and the display, and a web page only sees the last stages through a timestamp the browser assigns itself. This page measures dwell time and flight time, which are well-defined and genuinely useful, and says so rather than overclaiming.

What is key dwell time?

How long a key stays pressed, from keydown to keyup. It is a property of your typing rather than your keyboard. Efficient touch typists sit between 60 and 100 milliseconds; deliberate or hunt-and-peck typing runs well above 150.

What is flight time?

The gap between one key going down and the next going down — your typing rhythm. It maps closely to speed: roughly, WPM is 60000 divided by six times your median flight time in milliseconds.

Why report percentiles instead of an average?

Because an average hides inconsistency. Two typists with the same average dwell can behave very differently, one steady and one alternating between fast and slow. The gap between your median and your 95th percentile is what shows that.

How many keystrokes do I need to type?

At least thirty for the numbers to be meaningful, and a couple of hundred for the 95th percentile to be stable. Below thirty, the p95 is essentially just the second-slowest sample you produced.

What actually reduces keyboard input lag?

In rough order of impact: display refresh rate, any software in the input path such as remote desktop or remapping tools, USB polling rate, actuation distance, and firmware debounce time. The keyboard itself is usually a smaller factor than the display, and software in the path can outweigh all the hardware combined.

Is wireless slower than wired?

A modern 2.4GHz dongle adds roughly one to two milliseconds and is effectively indistinguishable in use. Bluetooth adds substantially more and varies with interference, so it is the one to avoid if responsiveness matters.

Should I aim for the shortest possible dwell time?

No. Below about 50 milliseconds you risk releasing before the switch fully registers, which shows up as dropped characters. Consistency matters more than speed here — a narrow gap between your median and p95 is the better goal.

Keyboard Layouts

Hardware Diagnostics

Individual Key Tests

Speed & Sound