How many Chinese characters do you need? 49 cover half of everything

By Carson 木林森 · September 2026

Across 73,073 characters of graded Chinese prose, the 49 most common characters account for half of everything on the page. 209 account for 80%. 841 account for 98%. The whole corpus uses only 1,380 distinct characters.

This is the question every beginner asks and almost every answer to it quotes a number with no corpus behind it. Below is a curve measured on a corpus you can read: the 145 passages in the reading library, HSK 1 through HSK 6. Only the Chinese in the passage bodies is counted.

The curve

The shape matters more than any single row. Getting from nothing to 80% of the text costs 209 characters. Getting from 80% to 98% costs 632 more. Getting the last 1% costs 384 beyond that. Effort per character rises the whole way, which is why a beginner improves fast and an intermediate learner feels stuck while working just as hard.

The first 100 characters, in order

Ranked by how often they occur. The last column is the share of all running text covered by everything down to and including that row.

Nothing in the top of that list is a thing. They are particles, pronouns, verbs of being and having, and the measure words and connectives that hold a sentence together. A beginner course that opens with animals and colours is teaching characters that sit far below this list, which is the same finding the prerequisite graph reaches from the other direction.

The order above is roughly the order GraphChinese teaches in. The course serves the next topic from what you have proven, and its ranking is built from how much each item unlocks rather than from a textbook chapter order. The whole course, HSK 0 through HSK 6, is $49 once for lifetime access.

How the curve moves with level

Each level's 98% figure is a fair target for reading that level comfortably. The jump from HSK 1 to HSK 6 is roughly threefold in distinct characters while the passages get several times longer, so later material repeats its own vocabulary harder than earlier material does. That is what makes extensive reading work at the top and not at the bottom.

The long tail is longer than the head

165 characters appear exactly once in the whole corpus. They are 12% of the distinct characters and 0.23% of the text. Studying to cover them is the least efficient thing a learner can do, and meeting them in context is the only sensible way to pick them up.

Characters on an HSK list cover 98.93% of the corpus. Only 56 distinct characters here sit on no HSK list. Read that as a statement about this material rather than about the language: these passages were written to HSK ceilings, so close agreement is designed in.

What this measurement does not tell you

• The corpus is graded, so the curve is flatter than native prose. Every passage was written to an HSK vocabulary ceiling. A novel, a news site or a group chat would need more characters for the same coverage, and the honest version of the headline is "49 characters cover half of this corpus", not "half of Chinese".
• 73,073 characters is a small corpus. Corpus linguistics works in millions. The number is printed here so you can weigh it rather than take the curve on trust.
• Coverage is not comprehension. 98% coverage still leaves one unknown character in every 50, which is several per paragraph, and an unknown character in the wrong place costs the whole sentence. Comfortable reading is usually put at 98% or better for a reason.
• Characters are not words. Knowing 好 and 意 and 思 does not give you 不好意思. The word-level version of this question is on how many words you need to speak.
• Recognising is not writing. Everything here counts characters seen on a page. The new HSK standard lists a separate, much shorter set for handwriting.

Questions about character counts

How many Chinese characters do you need to read?

On this corpus, 841 distinct characters cover 98% of the running text and 383 cover 90%. The whole 73,073-character corpus uses only 1,380 distinct characters. Reading unedited native material needs more, because a novel or a newspaper draws on names, technical words and idioms a graded passage avoids.

How many characters cover half of Chinese?

49. The 49 most common characters account for 50% of every character on the page. The curve is that steep at the start and that flat at the end: the last 1% of the text takes 384 more characters than the first 99%.

Is 2,000 characters enough for Chinese?

For this kind of material, comfortably. 1,380 distinct characters appear across all 145 passages, so 2,000 covers the whole corpus with room to spare. The number people quote for a newspaper is higher, and the useful question is not how many you know but which ones, since the 209 commonest do 80% of the work.

Do the HSK character lists cover real Chinese text?

Almost entirely, here. Characters on an HSK list account for 98.93% of this corpus, and only 56 distinct characters in the text sit on no HSK list at all. That is partly a property of the material: these passages were written to HSK vocabulary ceilings, so the agreement is expected rather than surprising.

Where does this measurement come from?

The 145 graded passages in the reading library, HSK 1 through HSK 6, counting only the Chinese in the passage bodies. It is recompiled from those files whenever the library changes, so the curve moves when the corpus does.

Related guides: The reading library · How to learn Chinese characters · How many words to speak Chinese · What depends on what in Chinese · The HSK levels, explained