DeviceBenchFree online typing speed test
Typing speed test scored in characters, not words
Type the passage for 1, 3 or 5 minutes and the page reports two speeds: gross words per minute, which counts every character you entered, and net, which takes a word off for each mistake still standing at the end. A word is five characters including spaces and punctuation, so a paragraph of long words scores the same as a paragraph of short ones. What a browser can see is the character your keyboard produced, never the switch that produced it — a key that drops a press appears here as a missing letter and as a slow row in the per-key table, not as a diagnosed fault.
- 100% free
- No signup
- 1, 3 or 5 minutes
- Gross and net WPM
- Per-key error table
Click the passage and start typing. Nothing arms the test in advance.
The first writing machines sold in any number were pitched as an aid for people who could not see the page, and only later found their real audience in offices. Early models hid the paper behind the carriage, so a typist could not read the line until it had scrolled into view. Makers argued for a decade about whether that mattered, and the market settled it: visible writing arrived, and nobody bought the old design again. A shipping forecast is read at the same pace every night, which
- entered
- 0
- standing correct
- 0
- still wrong
- 0
Nothing counted yet. Every character you enter is divided by 5 to reach a word, so the count never depends on where the spaces fall.
Filled in when the run ends: every character you hit at least twice, ranked by how often it came out wrong. — until then.
The clock counts only the time this field had focus, so answering a message mid-run costs you nothing. Pasting is refused and Enter is ignored, because both of them add characters that no finger typed. A phone works, but its keyboard predicts, corrects and spaces differently enough that the figure belongs in its own league rather than beside a desktop score.
How to run the test and read what comes back
Three steps, and the third is where the useful part is.
Choose the length before you touch a key
One minute is a sprint and reports what your hands can do at their best; five minutes reports what they do when concentration starts to slip, which for most people is a figure 10 to 20 WPM lower and the one that predicts a working day. Three minutes sits between the two and is long enough that a single fumbled word stops mattering. Changing the length loads a fresh passage, so you never retype prose you have already learned.
Click the passage and go — the first character is the start
Nothing arms the test beforehand, so the interval between deciding to begin and actually beginning is not charged to you. The character under the amber block is the one expected next; a wrong one turns red where it stands rather than stopping you, because a test that blocks on a mistake measures your willingness to correct rather than your speed. Click away and the clock stops until the field is focused again.
Read net first, then accuracy, then the key table
Net WPM is the number worth quoting because it has already paid for the mistakes still standing at the end. If it sits far below gross, the problem is accuracy and not speed, and slowing down by 10% usually raises net more than any amount of pushing does. The table underneath ranks the characters that went wrong most often with the mean gap before each one, which is where a specific finger, rather than a general pace, gives itself away.
Technical specifications
| Run lengths | 60, 180 and 300 seconds, on a passage of 5,200+ characters that cycles rather than running out |
|---|---|
| One word | 5 characters, spaces and punctuation included — the unit WPM was defined in |
| Gross WPM | Every character entered, divided by five, per minute of counted typing |
| Net WPM | Gross minus one word per minute for each error still standing when time runs out |
| Accuracy | Characters correct at the end over every character entered, so a typo you fixed still shows |
| Clock | Runs only while the typing field holds focus; paste, drop and Enter are refused |
| Per-key table | Up to 8 characters seen twice or more, ranked by errors, with the mean gap before each press |
| Data kept | None — the passage, the score and the key table are gone on reload |
Frequently asked questions
Why is my WPM here different from the one another site gave me?
Because there are three formulas in circulation and each site picks one without saying so. Gross WPM counts every character you entered and divides by five. Net WPM, the figure taught in typing instruction, subtracts one word per minute for every error left standing. A third convention counts only the characters that ended up correct, which lands between the two. This page shows gross and net together so you can see the size of the gap rather than wondering which number you were handed elsewhere.
Why does five characters count as one word?
Because words are not a fixed size and keystrokes are. The average English word runs about 4.7 letters, and adding the space that follows it puts the average keystroke cost of a word at roughly five — so the typewriter-era standard fixed a word at five characters and made scores comparable across any text. It is also why a passage full of long words does not lower your WPM here: the count is characters divided by five, not a tally of the words you finished.
Does using backspace hurt my score?
It costs accuracy and saves net WPM, which is usually the trade you want. Accuracy on this page is the characters that ended up correct over every character you entered, so a typo you fix still shows in the percentage — you did make it. Net WPM only penalizes errors that are still standing when time runs out, so correcting one buys back the word it would have cost. The one thing correcting does not buy back is the time you spent doing it, which is why heavy backspacing shows up as a low gross figure.
Why did the timer stop when I switched windows?
The clock only runs while the typing field has focus, so time spent in another tab or another application is not counted against you. This is not the usual behavior: most typing tests keep counting, which turns a ten-second interruption into a six-WPM penalty on a one-minute run. The trade-off is that the field must be focused for keystrokes to arrive at all, so the panel over the passage tells you when it is not.
Can I paste the passage instead of typing it?
No — paste and drop are both refused, and Enter is ignored. The measurement is keystrokes, and the clipboard does not produce any, so accepting one would mean reporting a speed nobody typed. Holding a key down is not blocked, though: the operating system repeats it about thirty times a second and every one of those characters counts as entered and as wrong, which is why leaning on the board collapses accuracy and net WPM together.
Would a mechanical keyboard make this number go up?
Barely, and not for the reason most people expect. A typical membrane key actuates around 2 mm of travel and a common mechanical switch around 2 mm as well, so the distance your finger covers is similar; what changes is how clearly you feel the actuation point, which affects confidence rather than raw speed. At 80 WPM you are producing roughly 6.7 characters a second, or one every 150 ms, and no switch on the market is the limiting factor at that rate. Where hardware does show up is a key that intermittently fails to register, and that appears here as a specific character with a high error count.
Is a score typed on a phone comparable to a desktop one?
No, and it is better to treat them as two separate records. A phone keyboard predicts, corrects and inserts spaces on your behalf, and even with autocorrect switched off in the field the touch target for each key is a fraction of a physical keycap. Thumb typing on glass peaks far below a trained touch typist on a full board, and comparing them tells you about the input method rather than about you.
About words per minute and why it is not a count of words
Words per minute is a keystroke rate wearing a friendlier name. A word, everywhere the unit is used seriously, is five characters — spaces and punctuation included — and that convention exists because words themselves are not a fixed size: English averages about 4.7 letters, so five keystrokes is what one costs once the space after it is counted. Fixing the unit is what makes two scores on two different passages comparable at all. The corollary catches people out: typing the word internationalization earns you four words, and typing a earns you two fifths of one.
The second thing a score depends on is what happens to mistakes, and here the field genuinely disagrees with itself. Typing instruction has used net WPM for a century — gross speed minus one word per minute for every uncorrected error — because a page that has to be retyped was not produced at any speed at all. Modern web tests more often count only the characters that came out right, which is gentler and produces a third number for the same run. Neither is wrong, but a figure quoted without saying which one it is cannot be compared with anything, which is why both appear on this page. Accuracy is measured separately, against every character entered, so correcting a mistake buys back the word without hiding that you made it. If it is raw hand speed rather than text you are curious about, the CPS test measures the same fingers with the passage taken away.
The per-key table is the part that turns a score into practice. Errors are attributed to the character that was expected, not to the one that arrived, because knowing you keep missing p tells you which finger is late while knowing you typed o instead tells you nothing you can drill. The space bar deserves its own note: it takes roughly one keystroke in six in English prose, more than any letter, and a worn one drops presses that read here as two words run together — the spacebar counter isolates that key on its own. Speed is also only one of the things a keyboard session asks of you: recalling a string long enough to type it is measured on the number memory test, while the two tests that share this page’s section measure how fast you respond to a cue at all — the reaction time test for a single press and the aim trainer for putting a pointer somewhere precise.
Where the text you type goes
Every number on this page is worked out by JavaScript running in the tab you are reading it in. Nothing you type, paste or open is uploaded, logged or kept, which is also why the tools carry on working after you disconnect from the network.
The passage is assembled inside the page from six paragraphs that ship with it, so nothing is fetched while you type and nothing you enter is ever joined back into a document. Characters are compared against the expected one, added to a counter and dropped. Reload and the run is gone — unlike the click tests on this site, this page writes nothing to local storage at all.