How many words do you need to read Chinese?
Characters get you further than words. We counted Chinese web text to show the gap.
We make Mandarin The Wild.
The 98% rule
Studies of second-language readers ask what share of a text's words a reader must know to follow it. The figure most often cited is 98%: one unknown word in every 50. It comes from a study of learners reading a short story in English, where most needed about 98% to understand it well on their own. Hu and Nation (2000), Unknown vocabulary density and reading comprehension, Reading in a Foreign Language 13(1).
Later work found no sudden threshold. Understanding rises steadily with the share of words known, and 98% remains a sensible aim. Schmitt, Jiang and Grabe (2011), The percentage of words known in a text and reading comprehension, Modern Language Journal 95(1). Another study put a minimum at 95%, one unknown word in 20, with 98% as the comfortable level. Laufer and Ravenhorst-Kalovski (2010), Lexical threshold revisited, Reading in a Foreign Language 22(1).
These studies are of English. We found none that sets a figure for Chinese, so take them as a guide, not a law.
Characters cover more than words
Chinese has a famous number: the 2,500 most common characters make up about 98% of the characters in everyday print, and 3,500 make up over 99%. List of Frequently Used Characters in Modern Chinese (1988), with its coverage figures.
We checked that on Chinese web text, counted with the engine behind our word cards. FineWeb-2, the web text we counted (its Mandarin part, cmn_Hani). Our count agrees: the 2,500 most common characters are 97.3% of the characters, and 96% of the words on a page are written only with them.
But knowing every character in a word doesn't tell you the word.
我买了很多东西。
thing; stuff
Also written 東西
你等一下，他马上就来。
right away; immediately
Also written 馬上
How many words a page takes
Here is the same web text counted by words. Each row says how much of it you could read if you knew that many of the most common words.
| The most common words you know | Share of the words on a page |
|---|---|
| 100 | 30% |
| 1,000 | 55% |
| 2,000 | 65% |
| 5,000 | 77% |
| 10,000 | 85% |
| 20,000 | 92% |
The 95% line comes at about 31,000 words, and 98% at more than 51,000. Web text is hard in a particular way. It is full of names, brands, places and topic words, each rare on its own, and together they are a large share of any page.
What the HSK levels cover
The HSK exam's word lists, from the 2025 syllabus in use since July 2026, run from 300 words at level 1 to about 5,400 at level 6. HSK: the official site of the Chinese Proficiency Test. Counted on the same web text:
| HSK words you know | Words | Share of the words on a page |
|---|---|---|
| Level 1 | 300 | 29% |
| Levels 1–3 | 1,000 | 42% |
| Levels 1–4 | 2,000 | 53% |
| Levels 1–6 | 5,400 | 69% |
| Levels 1–9 | about 11,000 | 74% |
That isn't a fault of the HSK. Its lists teach the words you need to speak and write, not the long tail of a news site. Our engine also cuts some phrases differently from the syllabus, which moves these shares a little.
What to do with this
- Read at your level. A story written for your level keeps you near 98%, so you follow it and pick up new words from context. The Library has graded stories from HSK 1, with recordings.
- Check a text before you read it. How hard is this text gives a text's HSK level and the words to learn before you read it.
- Look up the rest where they are. On a real page, a word you don't know is usually a name or a topic word. Point at it, and Mandarin The Wild shows its meaning in that sentence. See the extension.
- Learn words, not only characters. 东西 and 马上 show why. Keep the words you want with the line you met them in, and Review brings them back.
How we counted
We counted the Mandarin part of FineWeb-2, a large collection of web pages, cutting it into words with the engine behind our word cards. FineWeb-2, the web text we counted (its Mandarin part, cmn_Hani). Shares are of Chinese words only, leaving out numbers and Latin letters. Our list stops at about 107,000 different words, so the rarest are missing, and the true figures for 95% and 98% are a little higher than shown.
Questions
How many characters do I need to read Chinese?
About 2,500 cover 97% of the characters on a page. But knowing the characters is not knowing the words: 东西 means "stuff", not "east west".
Is HSK 6 enough to read the news?
Not on its own. On web text, HSK 1–6 covers about 69% of the words. The rest is mostly names and topic words, best learned by reading.
What share of words do I need to know?
About 98%, one unknown word in 50, for comfortable reading on your own. Studies of English readers put the minimum near 95%.
Sources
Checked on 13 October 2026. Prices and features change; tell us if one has.
- Hu and Nation (2000), Unknown vocabulary density and reading comprehension, Reading in a Foreign Language 13(1)
- Schmitt, Jiang and Grabe (2011), The percentage of words known in a text and reading comprehension, Modern Language Journal 95(1)
- Laufer and Ravenhorst-Kalovski (2010), Lexical threshold revisited, Reading in a Foreign Language 22(1)
- List of Frequently Used Characters in Modern Chinese (1988), with its coverage figures
- HSK: the official site of the Chinese Proficiency Test
- FineWeb-2, the web text we counted (its Mandarin part, cmn_Hani)