Test modes
Five ways to run a test and one way to practise. They are not interchangeable, and picking the wrong one is the most common reason a score looks wrong.
Updated 12 August 2026
The distinction that matters most. Sprint modes measure you on representative English. Forge trains you on text deliberately skewed toward your weak keys. A Forge score is not comparable to a Sprint score and is not meant to be — the text is harder on purpose.
Sprint modes
Time
Type for a fixed duration: 15, 30, 60 or 120 seconds. The text keeps coming; the clock decides when you stop. This is the default and the format nearly every published benchmark and employer test uses, so it is the mode to use when you want a number that means something to someone else.
Use 60 seconds unless you have a reason not to. Fifteen seconds is fun and almost meaningless — too short for a bad word to average out.
Words
Type a fixed count: 10, 25, 50 or 100 words. Everyone types the same amount of text, so the finish line is identical regardless of speed. Better than Time for comparing two attempts against each other, because the workload is held constant instead of the clock.
Quote
Real sentences rather than generated word lists. Punctuation, capitals and natural rhythm all appear, which makes it closer to actual writing — and usually a few WPM slower than word mode. If your Quote score is much lower than your Time score, punctuation and shift combinations are your bottleneck.
Zen
No timer, no word count, no end. Text generates until you stop. Use it to warm up, to practise without a countdown pushing you into errors, or to type for a while without producing a score you then have feelings about.
There is a page dedicated to this: typing practice with no time limit.
Custom
Paste your own text and type that. Useful for drilling a specific vocabulary — code, medical terms, names you keep misspelling — or for practising against the exact passage an exam uses.
Forge — adaptive practice
Forge is a separate view rather than another chip in the row, because it is doing something different. Instead of showing you representative English, it scores each key from your own history and generates words weighted toward your weakest ones.
The reason this is worth a separate mode: ordinary English is a poor teacher. Roughly 70% of keystrokes land on nine letters you almost certainly type well already, so general practice rehearses your strengths. Forge inverts the distribution.
- Letters unlock progressively, so a beginner is not thrown at the whole keyboard
- A newly unlocked letter starts at zero confidence and is therefore drilled hardest, immediately
- The generated words are real words, not random letters, so what you practise transfers
The mechanism is fully documented — confidence formula, unlock order, generation rules — at /docs/keyforge, and the evidence it is built on is at /docs/research.
Which to use
| If you want to… | Use |
|---|---|
| Get a number comparable to published benchmarks | Time, 60 seconds |
| Prepare for an employer or exam typing test | Time, matching their duration — often 1 or 10 minutes |
| Compare two attempts fairly | Words, fixed count |
| Practise realistic writing | Quote |
| Warm up, or type without being scored | Zen |
| Drill specific vocabulary | Custom |
| Actually get faster | Forge |
On comparing scores between sites. A mode label is not a guarantee of the same measurement. TypeCrt counts a word as five characters and scores at word level, which is documented with worked examples on the metrics page. When comparing against another site, check that it does the same before concluding one of them is wrong.
Start with 60 seconds
The default, and the one whose result you can actually place against published percentiles.
Take a 1-minute test →