How many words are in the Chinese language?
There is no single number, because Chinese is counted in two different units and most answers to this question don't say which one they mean. The biggest character dictionary ever compiled, 中华字海 Zhonghua Zihai, holds 85,568 characters. The biggest word dictionary, 汉语大词典 Hanyu Da Cidian, holds about 370,000 entries. A fluent adult knows a working slice of either. This page is about how big the language itself is — for how many of those words a learner needs, see how many words you need to speak Mandarin, which is a different question with a much smaller answer.
Characters and words are counted separately
Chinese keeps the two units apart even in its own vocabulary for them: a 字 zì is a character, and a 词 cí is a word — often two characters put together, like 学生 xuésheng (student) or 老师 lǎoshī (teacher), whose meaning isn't just the sum of its parts. Dictionaries split the same way. A 字典 zìdiǎn is a dictionary of characters, in the order they're built; a 词典 cídiǎn is a dictionary of words, which is what most people mean when they ask how big a language's vocabulary is. Any figure for "how many words are in Chinese" that doesn't say which kind of book it's counting from isn't worth repeating.
The numbers, side by side
| Reference | Published | Size | What it counts |
|---|---|---|---|
| Kangxi Dictionary 康熙字典 | 1716 | 47,043 | characters |
| Table of General Standard Chinese Characters 通用规范汉字表 | 2013 | 8,105 | characters China designates as standard, in three tiers |
| Zhonghua Zihai 中华字海 | 1994 | 85,568 | characters — the most ever catalogued in one dictionary |
| Xiandai Hanyu Cidian 现代汉语词典 (7th edition) | 2016 | about 70,000 | words and set phrases — the dictionary most households own |
| Hanyu Da Cidian 汉语大词典 | 1986–1994 | about 370,000 | words and phrases — the largest ever compiled |
| HSK 3.0 advanced band, GF 0025-2021 | 2021 | 11,092 | words a non-native learner is expected to know at the top level |
Six answers to one question, none of them wrong. What follows is where each comes from.
The character dictionaries, three centuries apart
The Kangxi Dictionary, commissioned by the Qing emperor and published in 1716, held 47,043 character entries and set the standard reference point for Chinese characters for the better part of three centuries. Zhonghua Zihai, compiled in 1994, holds 85,568 — nearly double, and the largest character count any single dictionary has catalogued.
Neither is a ceiling. Unicode, the digital standard that gives every character its own code point so computers can display it, had identified more than 100,000 Chinese characters by 2025. Most of those are not words anyone needs for ordinary reading — historical variants, rare surnames, characters that appear once in a particular classical text — they are there so that nothing gets lost, not because anyone needs to learn them.
China's own working list: 8,105 characters in three tiers
The practical answer, if you want one list rather than a historical maximum, is the Table of General Standard Chinese Characters (通用规范汉字表), issued by the State Council in 2013 to replace the older standards. It holds 8,105 characters in three tiers: 3,500 for everyday general use, a further 3,000 for publishing, dictionary-making and information processing, and a last 1,605 covering surnames, place names and technical or classical terms — numbers precise enough that a page can quote them without hedging.
Measured independently — by how often a character actually turns up, rather than by which official tier it sits in — our own count across 85 million words of Chinese film and television dialogue finds that the top 3,500 characters by frequency already cover 99.4% of everything written or said in that corpus. Two different ways of drawing the line, by designation and by use, land on almost the same size.
The word dictionaries: from the household shelf to the largest ever made
现代汉语词典 Xiandai Hanyu Cidian is the dictionary most Chinese households actually own, published by the Commercial Press and now in its seventh edition (2016), with about 70,000 entries — characters, words, set phrases and idioms together. It has grown steadily rather than suddenly: 56,000 entries in its first edition in 1978, 60,000 by 1996, 65,000 by 2005, 70,000 by 2016.
汉语大词典 Hanyu Da Cidian is a different kind of book entirely — closer to a historical dictionary on the scale of the OED than to something kept on a desk. More than 300 scholars, led by chief editor Luo Zhufeng, spent from 1979 to 1994 compiling it: thirteen volumes, over 23,000 head characters, roughly 370,000 words and phrases drawn from 1.5 million citations running from classical to modern Chinese. It records words a classical text used once, not words anyone needs to speak the language today.
What a fluent adult actually knows
None of the above is what's in anyone's head. China's own functional-literacy standards, set in 1988 and unchanged since, ask for 1,500 characters from farmers and 2,000 from urban workers and officials. About 6,500 characters are in regular modern use at all, and the first 3,500 of those — the same size as Tier 1 of the 2013 standard list — are enough to write 99% of what appears in ordinary reading material. The average Chinese college student is estimated to know around 5,150 characters, which is enough to write more than 30,000 words.
Put the two halves of this page together and the gap is the real story: the biggest dictionary ever made holds 370,000 words, and a well-read adult operates on something like a tenth of that, built from a few thousand characters recombined over and over.
What a foreign learner is held to: 11,092 words
China's own standard for teaching Chinese to non-native speakers, the Chinese Proficiency Grading Standards for International Chinese Language Education (GF 0025-2021) — the basis of the "HSK 3.0" exams — puts a number on fluency from the other direction. Its top band, levels 7 to 9, asks for a cumulative 11,092 words, 3,000 characters you can recognise and only 1,200 you can write by hand. That is a small fraction of the household dictionary's 70,000 entries, and it is explicitly a learner's target, not a native speaker's — the gap between this and real fluency, level by level, is worked through in how long it takes to learn Mandarin.
What our own numbers say
Word counts depend heavily on who is doing the counting, and ours is honest about the method. Across the same 85-million-word corpus of film and television dialogue, automatic segmentation turns up 766,612 distinct word-strings — more than even Hanyu Da Cidian's 370,000 entries. That is not a dictionary-quality count: a segmenter counts a name, a number or a one-off foreign transliteration as its own "word" just as readily as it counts a real one — the same measurement problem that makes a learner's frequency list depend on the segmenter, not just the corpus. The same corpus, split into characters instead, turns up only 11,302 distinct characters — a small fraction of Zhonghua Zihai's 85,568, and proof of how concentrated real usage is next to what a dictionary has to hold.
There's a second way our numbers show the size of Mandarin's vocabulary: measured on equal-sized samples of text, Mandarin needs its 10,000 most common words to cover 88.6% of ordinary dialogue, while Spanish reaches that same coverage with only 5,607 words. The hardest languages to learn sets that comparison out for six languages at once; the short version is that Mandarin's working vocabulary for everyday speech is unusually large well before you get anywhere near a dictionary's hundreds of thousands.
So, how many words are there, really?
- If you mean characters: a specialist dictionary holds 85,568; China's own working standard holds 8,105; ordinary reading needs about 3,500 of them.
- If you mean words: the dictionary on a household shelf holds about 70,000; the largest ever compiled holds about 370,000; nobody uses anywhere near either.
- If you mean "how many do I need to learn": that's a smaller and much more answerable question, covered on how many words you need to speak Mandarin and is Mandarin hard to learn — and the part of the language that makes every one of those words hard to say correctly, the four tones, has its own page: the Mandarin tones explained.
Verbavia (our own course) teaches Mandarin — the first language of 990 million people — from a real frequency list rather than a simplified teaching vocabulary, with every word's characters and audio together. The Mandarin course is free on the web in any browser including on a phone, the 1,000 most common words are free to browse, and the two-minute vocabulary test estimates how many of them you already know.
Sources
- Wikipedia: Kangxi Dictionary — 47,043 character entries, published 1716.
- Wikipedia: Zhonghua Zihai — 85,568 characters, compiled 1994.
- Wikipedia: Table of General Standard Chinese Characters — 8,105 characters, issued 2013; Chinese Wikipedia: 通用规范汉字表 — the exact tier counts, 3,500 / 3,000 / 1,605.
- Wikipedia: Xiandai Hanyu Cidian — about 70,000 entries in the 7th edition (2016), Commercial Press; edition history from 56,000 in 1978.
- Wikipedia: Hanyu Da Cidian — about 370,000 words and phrases, 13 volumes, compiled 1979–1994 under chief editor Luo Zhufeng.
- Wikipedia: Chinese characters — more than 100,000 characters catalogued in Unicode as of 2025.
- Wikipedia: Literacy in China — the 1988 functional-literacy standards (1,500 / 2,000 characters); about 6,500 characters in regular use, with 3,500 covering 99% of popular reading material; the college-student average of about 5,150 characters.
- Wikipedia: Hanyu Shuiping Kaoshi — the Chinese Proficiency Grading Standards for International Chinese Language Education (GF 0025-2021) and its cumulative 11,092-word advanced band.
- Lison, P. & Tiedemann, J. (2016). OpenSubtitles2016: Extracting Large Parallel Corpora from Movie and TV Subtitles. Proceedings of LREC 2016. aclanthology.org
- Dave, H. FrequencyWords: frequency lists from OpenSubtitles 2018 (
zh_cn), content licensed CC BY-SA 4.0. github.com. The word-type, character-type and coverage figures are computed byscripts/frequency_curves.pyand printed from its output rather than typed.