Device Bench

Device test · Input

Typing Test

Type the words shown — WPM, accuracy and consistency come after.

Type the words above. The clock starts with your first counted character and stops at the chosen time. Space submits a word. Under 5 seconds, no WPM is shown.

WPM— Raw WPM— Accuracy— Consistency— Time60.0 s left

The stream uses a 300-word list written for this page. The first 120 entries have three times the draw weight of each of the remaining 180. Use the same “Stream seed” and time mode to repeat a word stream; “New words” picks a fresh seed.

On a phone or tablet, autocorrect, predictive text and swipe entry can insert or replace text without a separate keypress for each character. The test does not block these features or adjust the score for them, so touch-keyboard results are not comparable with hardware-keyboard results.

No result yet. Let the timer finish or use “Stop and score” to end early. Runs stopped before 5 seconds show no WPM or consistency.

WPM (correct characters ÷ 5 ÷ minutes)—
Raw WPM (counted characters ÷ 5 ÷ minutes)—
Accuracy (correct ÷ all counted)—
Consistency (CV of the per-second speeds)—
Words submitted—
Characters counted—

How this test works

The readout shows four typing measures and the time remaining, or elapsed time after the run. WPM is correct characters ÷ 5 ÷ minutes; raw WPM is every counted character ÷ 5 ÷ minutes. Here, a “word” means five characters, not one word from the stream. Accuracy is correct characters ÷ every counted character, shown as a percentage. Consistency is the coefficient of variation of the per-second speeds, also shown as a percentage; lower means a steadier pace across seconds. A counted letter is correct when it matches the target at that position. The submitting space is correct only if the whole word matches. Extra letters past the end of a word count as wrong.

The timer uses performance.now(), a monotonic clock, and starts with your first counted character. Spaces before a word or immediately after a submitting space are ignored and do not start it. A full run lasts 15, 30, 60, 180 or 300 seconds; input arriving at or after the time limit is not counted. “Stop and score” uses the elapsed time, capped at the selected length. Character counts are based on increases in the input field’s length: backspace is never counted and never un-counts. Deleting a mistake does not remove its earlier count, and existing letters shifted by an edit are not counted again. A replacement that leaves the field the same length adds no counted characters. For consistency, each completed second contributes one sample: correct characters in that second ÷ 5 × 60. Characters are grouped by their arrival time. Consistency is the population standard deviation of those samples divided by their mean, multiplied by 100. It needs at least two completed seconds and a nonzero mean; the final partial second is excluded.

WPM depends on the difficulty of the text. This stream uses 300 words of three to five letters, without capitals, punctuation or numbers. Keep the time mode, seed and keyboard settings the same when comparing runs. Touch-keyboard features can change text without a separate keypress for each character, and the score does not compensate; those results are not comparable with hardware-keyboard results. If you stop before five seconds, the result shows no WPM or consistency, but character counts and accuracy remain available. A full 15-second run provides 15 consistency samples. The word stream is visual and hidden from screen readers, with no spoken or audible alternative, so you need to see it to take this test. The result measures this timed copying task, not the accuracy of your own writing or the condition of your keyboard.

Frequently asked questions

How is the WPM figure worked out?
WPM is correct characters ÷ 5 ÷ elapsed minutes. The five-character unit lets the result use words per minute even though the stream’s words vary in length. A counted letter is correct when it matches the target in that position as it arrives; the space that submits a word is correct only if that word is complete and correct. Extra letters beyond the target word count as wrong. Backspace adds nothing to the count and does not remove earlier characters or mistakes. When an edit shifts existing letters, those letters are not counted again. Raw WPM uses all counted characters, right or wrong, with the same divisor.
Why is my raw speed higher than my WPM?
Raw WPM includes every counted character; WPM includes only correct ones. Both use the same elapsed time and five-character word unit, so wrong characters create the gap. These include extra letters and the submitting space after an incomplete or incorrect word. Leading spaces and repeated spaces that submit nothing are ignored. Accuracy expresses the same counts as a percentage: correct ÷ all counted × 100. Deleting a mistake does not erase it from these totals.
Why does pasting text into the box do nothing?
The test ignores pasted and dropped text and shows a notice under the input box. Neither action inserts text, adds characters to the score or starts or resets the timer. Autocorrect and other touch-keyboard replacements are not blocked. Counts are based on net increases in the input field’s length, so a replacement of the same length adds no counted characters.
What does the consistency figure mean?
Consistency is the coefficient of variation (CV) of the completed one-second speed samples, shown as a percentage. Each sample is correct characters ÷ 5 × 60 for that second. The calculation is population standard deviation ÷ mean × 100. Lower values mean less variation between seconds; bursts and pauses tend to raise it. Characters are assigned to the second they arrive in, rather than the next animation frame. The final partial second is not sampled. CV needs at least two completed seconds and a nonzero mean; the results panel also withholds it for runs under five seconds.
Are phone and hardware-keyboard results comparable?
On a phone or tablet, autocorrect, predictive text and swipe entry can insert or replace text without a separate keypress for each character. The test does not block these features or adjust the score for them, so touch-keyboard results are not comparable with hardware-keyboard results. Keep the time mode, stream seed, device and keyboard settings the same when comparing repeated runs. Text difficulty matters too: this word stream has no capitals, punctuation or numbers, so scores from a different typing test need not match.

Related tests and guides