Free Typing Speed Test
Measure your WPM, accuracy, and consistency — no login required
Most typing tests give you a number and no way to check it. This one shows its working: what you will be typing, how net WPM is derived from your keystrokes, how the consistency score and the key heatmap are computed, and where the percentile on your result card actually comes from.
What you will actually be typing
In passage mode the test draws from a corpus of 30 original passages — 4,598 words in total, averaging 153 words each — written as ordinary general-interest prose rather than lifted from anywhere. Three of them are below, one from each difficulty level, with the measured properties of that specific text underneath.
The sun rises in the east and sets in the west. It gives us light and warmth every day. Without the sun, life on Earth would not be possible. Plants need sunlight to grow, and animals depend on plants for food. Humans also rely on sunlight for energy and daily activities. The sun helps us tell time and plan our day. In the morning, the sunlight feels soft and fresh. It slowly becomes stronger as the day goes on. In the evening, the sky changes color and becomes calm and …
Communication is one of the most essential skills in modern life, shaping how individuals interact, collaborate, and build relationships across different environments. Every day, people exchange ideas through conversations, emails, messages, and presentations. Clear communication helps reduce confusion, improve efficiency, and strengthen trust between individuals. In professional settings, effective communication ensures that tasks are understood correctly and executed without unnecessary delays. It also plays a key role in leadership, where the ability to explain goals and expectations can directly influence team performance. Listening is an equally important part of communication …
The rapid advancement of technology has transformed nearly every aspect of human life, from communication and education to business and entertainment. Digital tools have made it possible to access information instantly, connect with people across the globe, and perform complex tasks with ease. This transformation has increased efficiency and created new opportunities, but it has also introduced new challenges that require careful consideration. One of the most significant concerns is data privacy, as individuals share large amounts of personal information online. Protecting this data has become a priority for both …
In word mode you get something quite different: a stream of individual words with no punctuation and no capitals, generated on the spot from a fixed pool. There is nothing to read and nothing to lose your place in — just the words themselves, one after another, which is why it measures something narrower than the passages above do.
Who this test is for
Anyone who wants an honest number rather than a flattering one. That covers three groups in particular: people who have never measured their speed and want a baseline; people preparing for a specific assessment — a government exam, a BPO screening, an office typing test — who need to know how far off the pace they are; and people already practising who want to see whether last month made a difference.
It is a general test, so it is the right starting point but rarely the right finishing point. If you are preparing for something specific, take a baseline here and then move to the page for that exam or role, where the passages match the text you will actually be given and the scoring matches the rules you will be marked under.
| Durations | 30 seconds, 1 minute, 2 minutes or 5 minutes |
|---|---|
| Modes | Word mode (individual words) and passage mode (continuous prose) |
| Display | Live typing (type over the text) or classic (separate input box) |
| Scoring formula | 5-character standard word; one standard word deducted per incorrect word |
| Backspace | On by default, can be turned off |
| Difficulty levels | Easy, Medium and Hard — measured differences below |
| Passage corpus | 30 passages, 4,598 words, averaging 153 words each |
| Word pool | Generated continuously — the queue never runs out mid-test |
| Report card | Net and gross WPM, accuracy, consistency, mistyped keys, percentile estimate |
| Cost | Free, no login, results never leave your browser |
Word mode and passage mode measure different things
The toggle at the top of the test is not a cosmetic choice. The two modes compare what you typed against the target in genuinely different ways, and the difference shows up most sharply in what happens when you make one small mistake.
| Word mode | Passage mode | |
|---|---|---|
| What is on screen | A stream of individual words with no punctuation or capitals | A continuous paragraph, punctuation and capitals included |
| How you advance | Press space to commit a word and move to the next | Type straight through; the cursor follows your character count |
| How a word is judged | Against the word at the same position in the queue, on its own | Against the characters at the same positions in the paragraph |
| Effect of one dropped letter | Costs you that word only. The next word starts clean | Shifts every following character by one, so the rest reads as wrong |
| What it is good for | Raw finger speed, warm-ups, drilling problem keys | Exam and assessment conditions, sustained accuracy, rhythm |
That fourth row is the one worth understanding, because it surprises people. Passage mode checks the character you typed at position 300 against character 300 of the passage. Miss a single letter early on and every character after it sits one place to the left of where the comparison expects it — so the screen fills with red even though your fingers are typing exactly the right thing. Nothing has gone wrong with the test; you have simply lost alignment, and backspacing to the point where you dropped the letter restores it.
Word mode cannot do this. Each word is compared with the word at the same position in the queue and nothing else, so a mistake is contained: you lose that word and the next one starts clean. That containment is exactly why word mode is better for drilling and worse as a simulation of a real assessment, where losing your place is part of what is being tested.
How your score is calculated
Typing speed is not measured in real words, because real words vary in length and a test full of short ones would flatter you. The convention, used here and by every other test on this site, is the standard word: five keystrokes, spaces included. Every figure on the result card is built from that unit.
Worked through with real numbers. Suppose you take the 5-minute passage test, type 1,043 characters, and nine of your words come out wrong:
| Characters typed, spaces included | 1,043 |
|---|---|
| Standard words — 1,043 ÷ 5 | 208.6 |
| Gross WPM — 208.6 ÷ 5 minutes | 41.7 → 42 |
| Words typed incorrectly | 9 |
| Net standard words — 208.6 − 9 | 199.6 |
| Net WPM — 199.6 ÷ 5 minutes | 39.9 → 40 |
Two things follow from this that are worth holding on to. First, the deduction is per wrong word, not per wrong character. A word with three wrong letters costs you exactly the same as a word with one, which stops a single dropped letter from being punished several times over. Second, accuracy is judged whole-word and without partial credit: a word counts as correct only when every character in it matches, so 95% accuracy means one word in twenty was wrong somewhere, not that 95% of your keystrokes landed.
Gross WPM is reported alongside net for diagnosis. If the two are close, your accuracy is fine and any gains have to come from speed. If there is a wide gap, you are already fast enough and the whole of your improvement is sitting in the error column.
Where your score sits, and how we estimate that
The result card tells you roughly what share of typists you are faster than. Here is exactly where that comes from, because it is an estimate and you should know its limits.
Nothing you type on this site is uploaded — scoring runs entirely in your browser — which means we have no visitor data to rank you against and will not pretend otherwise. Instead the percentile is read off a model. We treat typing speed as normally distributed with a mean of 52 WPM and a standard deviation of 20.2, and both numbers come from the largest published study of the question: an analysis of 136 million keystrokes from 168,000 volunteers, which reported an average of 51.56 WPM with a standard deviation of 20.2.
The honest caveat is in the same paper. It states plainly that the observed distribution is not normal — it carries a positive skew of 0.513, meaning there is a longer tail of fast typists than a bell curve allows for. Our model therefore places you reasonably well near the middle and progressively less well towards the edges, and it will tend to overstate how unusual a very high score is. It is a placement, not a measurement, and the result card describes it as an estimate for that reason.
Two more limits worth stating. The study population was self-selected — people who chose to take an online typing test, 72% of whom had taken a typing course — so it skews faster than the general public. And it measured with its own transcription task, not with ours. For what the numbers mean in practice rather than in a distribution, the guide to what counts as a good typing speed goes into it properly.
How the consistency score is computed
Consistency is the figure most people ignore and the one that best predicts whether a score will hold up under pressure. It is our own measure, so here is the method in full.
While the test runs, the engine recalculates your net standard-word total every second — total keystrokes divided by five, minus one for each word you have got wrong so far. Every ten seconds it takes the difference between that total and the total at the previous checkpoint, multiplies by six to express it as a per-minute rate, and stores it. Those stored figures are your speed buckets.
At the end it takes the mean of the buckets and their standard deviation, divides the second by the first to get a coefficient of variation, and subtracts that percentage from 100. A run whose buckets were 48, 51, 49 and 50 WPM barely varies and scores in the nineties. A run of 62, 44, 58 and 39 has the same average but scores far lower.
| Duration | 10-second buckets | Best used for | Where to be careful |
|---|---|---|---|
| 30 seconds | 3 | A warm-up, or a quick check after changing something about how you type | Three consistency buckets is too few to mean much. Treat the rhythm figure as noise |
| 1 minute | 6 | The default benchmark. Long enough to be honest, short enough to repeat | Still short enough that one bad patch moves the score several WPM |
| 2 minutes | 12 | The shortest run where the consistency score starts to describe you rather than the run | Most people lose 2–4 WPM against their 1-minute figure. That drop is the point |
| 5 minutes | 30 | Stamina, and preparation for the 10 and 15-minute exam formats | Passage mode moves to the next passage when you finish one, so the topic will change mid-run |
The limitation is arithmetic rather than opinion: a standard deviation over three values is not a stable quantity. On the 30-second test the consistency figure will jump around between runs for reasons that have nothing to do with your typing. From two minutes upward it starts describing you. And because the buckets are cumulative differences, a burst of errors shows up as a low bucket rather than a fast one — the measure tracks useful output, not finger movement.
How the mistyped-key list is derived
After the test, the report card lists the characters you got wrong most often and what you hit instead. The derivation is simple: for every position where the character you typed differs from the character expected there, the engine records the pair — expected character, typed character — and counts it. At the end it groups by expected character, finds the substitute you produced most often for each, and shows the six with the highest totals.
What makes this useful is that mistyping is not random. Most people have two or three keys that account for a large share of their errors, usually reaches that their weakest fingers make badly, and seeing the substitution tells you which. Consistently typing e where r belongs is a different problem from typing r where t belongs, and the fix is different too.
One caveat, and it matters. Because the comparison is positional, a dropped or extra character in passage mode shifts everything after it and generates a long list of substitutions that are really one alignment error wearing a disguise. If a single run produces a heatmap that looks like noise, check whether you lost your place partway through — and read the list from a run where you did not. Word mode is immune to this, which makes it the better mode for diagnosing key problems.
Reading your report card
Six figures appear when a test ends. Only one of them is your score; the rest exist to tell you what to do about it.
| Figure | What it is | What to take from it |
|---|---|---|
| Net WPM | Your score. Total keystrokes ÷ 5 ÷ minutes, minus one word for every word you got wrong | This is the number an employer or exam reports. Everything else is diagnosis |
| Gross WPM | The same calculation with no deduction — your raw speed ceiling | A large gap between gross and net is an accuracy problem, not a speed problem |
| Accuracy | The share of words you completed exactly right, character for character | A word with one wrong letter counts as wrong. There is no partial credit |
| Consistency | 100 minus the coefficient of variation of your 10-second speed buckets | Below about 70% usually means you are starting fast and fading, not that you are erratic |
| Mistyped keys | The characters you got wrong most often, with the key you hit instead | Two or three keys normally account for most errors. Those are what to drill |
| Rank and percentile | Where your net WPM sits in a modelled population — see the methodology below | An estimate from a published average, not a measurement of this site’s visitors |
The card also compares the run with your previous one, which is stored in your browser and nowhere else. That comparison is the most useful thing on the page over time and the least useful in any single session — day-to-day variation of several WPM is normal, and reading meaning into one run against another will mostly mislead you. Compare weeks, not attempts.
What the three difficulty levels actually change
Difficulty here is a property of the text, not of the scoring, and we measure it rather than assert it. These figures are computed from the corpus itself by scripts/corpus-stats.mjs, which runs on every deploy.
| Level | Avg word length | Words 10+ chars | Avg sentence | Passages |
|---|---|---|---|---|
| Easy | 4.76 | 3.8% | 10 words | 10 |
| Medium | 6 | 15.7% | 15.5 words | 10 |
| Hard | 6.05 | 17.2% | 16.1 words | 10 |
The step from Easy to Medium is large and real: average word length rises from 4.76 to 6 characters, the share of words ten characters or longer goes from 3.8% to 15.7%, and sentences grow from 10 words to 15.5. Expect to lose real speed crossing it.
The step from Medium to Hard is not. On our own measurements the two sets are close to indistinguishable — 15.7% against 17.2% long words, average word length 6 against 6.05. We would rather publish that than dress it up: this corpus is on the list to be rebuilt so that Hard is genuinely harder, and until it is, treat Medium and Hard as one level with two names. The professional and exam corpora on this site do carry a real three-step ladder, and their pages show the same table.
What to practise, based on what the report card said
A general test is a diagnosis, not a treatment. Once you know where the loss is, the drill for it is somewhere more specific.
| If this is the problem | Go here | Why |
|---|---|---|
| Digits break your rhythm | Numbers test | Isolates the number row so the reach stops costing you a beat mid-word |
| Shifted characters slow you down | Symbols test | Drills the shift combinations that appear in code, passwords and references |
| Mixed letter-and-digit strings | Alphanumeric test | Order IDs, policy numbers and reference codes, where nothing can be guessed |
| Under about 30 WPM, or you look at the keyboard | Learn Typing course | Fixes finger placement first. Speed work before that just entrenches the habit |
| You want the technique, not another test | How to type faster | The practice method itself, written once and kept in one place |
If you are preparing for a specific exam, go to that exam’s page rather than practising here — the passages, the duration and the backspace rule all change, and practising under the wrong rules is worse than not practising. The full list of tests covers government exams, corporate screenings, professional roles and programming languages.
Frequently asked questions
Why is my net WPM lower than my gross WPM?
Gross WPM is raw output: every keystroke you made, divided by five, divided by the minutes elapsed. From that raw total, one standard word is taken away for every word that came out wrong. Type at 60 gross WPM for a minute with eight bad words in it and you finish on 52 net. Net is the figure that matters — it is what typing assessments and government exams report, and it is the one shown largest on the result card.
Does a single mistyped letter cost me a whole word?
Yes, and that is deliberate. Nothing in a word earns partial credit — get one letter wrong and the whole thing is scored wrong, which takes a standard word off your net total. Every other test on this site applies the same rule. The alternative — counting each wrong character separately — punishes a single slip several times over, because one dropped letter throws off every character after it.
Why did one missing character turn the rest of the passage red?
Passage mode compares by position: the character you typed 300 keystrokes in is checked against character 300 of the passage. Drop one letter early and everything after it sits one place to the left, so the comparison fails on almost every character even though your fingers are typing the right thing. Use backspace to get back into alignment as soon as you notice. Word mode does not behave this way — each word is judged on its own, so a slip costs you one word and the next starts clean.
What does the consistency score actually measure?
Every ten seconds the test records how many net standard words you produced during that bucket and converts it to a WPM figure. At the end it takes the standard deviation of those buckets, divides by their mean to get a coefficient of variation, and subtracts that from 100. A score of 90 means your speed barely moved across the run. A score of 55 usually means you opened fast and faded, which is much more common than genuinely erratic typing.
Is the percentile on my result card based on real users?
No, and we would rather say so than imply otherwise. Nothing you type here is uploaded, so we have no visitor data to rank you against. The percentile comes from a model: a normal distribution centred on 52 WPM with a standard deviation of 20.2, both taken from a published study of 168,000 typists. That study also notes the real distribution is positively skewed rather than normal, so treat the percentile as a rough placement, not a measurement.
Which duration gives the most reliable number?
Two minutes is the shortest run that describes you rather than the run. Thirty seconds and one minute both flatter people who can sprint, and the consistency score at 30 seconds is built from only three buckets, which is not enough to be meaningful. If you are preparing for an exam with a 10 or 15-minute passage, work up to the 5-minute test — the drop between your 1-minute and 5-minute figures is the stamina gap you need to close.
Should I use word mode or passage mode?
Word mode for raw speed and for drilling. It strips out punctuation, capitals and the risk of losing your place, so it measures finger speed almost in isolation. Passage mode for anything that resembles a real assessment: continuous prose with punctuation, where losing alignment has consequences and you have to sustain attention. Most exam and employer tests are passage tests, so if you are preparing for one, practise in passage mode.
Can I run out of words on a long test?
No. Word mode tops the queue up while you type, so it stays ahead of you however fast you go — this was a real defect until September 2026, when anyone above roughly 50 WPM could type off the end of a fixed list on the 5-minute test and see their score collapse. Passage mode moves to the next passage in the set when you finish one, and cycles back to the start if you get through them all.
Why is backspace enabled by default when exams disable it?
Because most people are here to measure and improve, not to sit a mock exam, and correcting a slip is what you would do in real work. You can turn it off in the settings bar, and you should before an exam that forbids it — SSC, CPCT and several High Court tests all lock the key, and first-pass accuracy is a different skill from corrected accuracy. Our government exam pages run with it disabled by default for that reason.
Are my results stored or sent anywhere?
Scoring happens entirely in your browser. Your most recent result is kept in your browser’s local storage so the next run can show you the change, and it never leaves your device. There is no account, no login, and no upload of what you typed. The trade-off is that we cannot show you real cross-user statistics — see the percentile question above.
Sources and method
Corpus figures — passage counts, word counts, average word length, long-word share and sentence length — are computed from our own passage files by scripts/corpus-stats.mjs and regenerated on every deploy, so they cannot go stale. The scoring, consistency and heatmap methods described above are descriptions of what the code on this page does, not claims about typing in general.
The average of 52 WPM and standard deviation of 20.2 behind the percentile estimate are taken directly from the study below, which reports an average of 51.56 WPM (SD 20.2) across 168,000 participants and states that the distribution is positively skewed. We have not rounded, adjusted or re-derived those figures, and we have taken them from the paper rather than from a secondary summary.
- Dhakal, Feit, Kristensson & Oulasvirta — "Observations on Typing from 136 Million Keystrokes"ACM CHI 2018 · Aalto University and University of Cambridge · link checked 3 Sept 2026