Ask this question anywhere and you get four answers. Two thousand. Three thousand. Eight thousand. "Well, it depends." All of them are defensible, which is exactly why none of them help.

The problem is that "how many characters do I need" is really three questions wearing one coat. How many exist? How many does a literate adult know? And how many do I need before Chinese stops feeling like a wall? Only the third one is your question, and it is the one nobody answers with actual numbers.

So I went and counted. What follows is measured directly against the HSK 3.0 character and vocabulary lists, the same standard our own HSK character lists are built on, and it shows exactly what each milestone unlocks. One of the numbers genuinely surprised me.

The short answer

If you want a single number to aim at, aim at 1,500. That is roughly where Chinese stops being a decoding exercise and starts being a language you read, slowly, with a dictionary nearby.

If you want the honest version, it depends on what you want to do with it. Around 300 characters, you can read your first real things: signs, menu items, short messages. Around 900, more than half of the standard vocabulary list opens up and graded readers become genuinely readable. Around 1,500 to 1,800, you are functional for most everyday text, with lookups. Around 3,000, you are in the working range of a literate adult and can read news and novels without constant interruption. Past that, you are into specialist and classical characters, and the returns fall off a cliff.

A large dictionary lists tens of thousands of characters. Nobody knows them all, native speakers included, and you can safely stop worrying about that number forever.

Why every source gives you a different figure

Two reasons, and both are worth understanding before you trust any number you read, including mine.

The first is that characters are not words. This is the single biggest source of confusion in the whole topic. 电 means electric and 脑 means brain, but the word everyone actually uses, 电脑, means computer. Modern Chinese runs on multi-character words. The HSK 3.0 word list runs to about 11,000 entries across its nine levels, and of the 10,943 distinct words in it, 8,355 are exactly two characters long. That is three quarters of the entire list built from pairs. So a source counting words and a source counting characters can describe identical knowledge with numbers that differ by a factor of three. Characters are the alphabet. Words are the language.

The second is that most quoted figures describe native literacy, not your goal. What a Chinese adult knows after twelve years of schooling conducted in the language is a fact about somebody else. It is not a useful target for someone who wants to read a menu next spring.

The number that actually decides whether you can read something

Here is the concept that replaces character counts entirely, once you have it: coverage. Not how many characters you know in total, but what percentage of the characters on the page in front of you are ones you know.

Chinese frequency is brutally lopsided, which is what makes this work in your favour. The hundred most common characters alone account for something close to half of everything you will ever read, and the top thousand get you to roughly ninety percent.

Ninety percent sounds like victory. It is not. At ninety percent, one character in ten is unknown, which is two or three per line. That is not reading. That is decoding with a dictionary open.

The thresholds that matter are higher, and closer together, than people expect. At 95 percent coverage, roughly one unknown character every twenty, reading is comfortable: you can guess from context and keep moving. At 98 percent, roughly one unknown every fifty or about one per two lines, reading becomes independent and a book stops being work.

Almost all of your character learning goes into that gap between 90 and 98 percent. It is also why progress feels fast at the beginning and slow in the middle. The early characters are the ones you meet constantly.

What each milestone actually unlocks, measured

This is the part I could not find anywhere, so I calculated it. I took the HSK 3.0 vocabulary list, all 10,943 distinct words, and checked each one against the character bands we publish on our HSK level pages, asking a single question at every level: how many of these words are written using only characters you have already learned?

300 characters, the HSK 1 set: 1,399 words, or 12.8 percent.

600 characters, through HSK 2: 3,484 words, or 31.8 percent.

900 characters, through HSK 3: 5,685 words, or 52.0 percent.

1,200 characters, through HSK 4: 7,202 words, or 65.8 percent.

1,500 characters, through HSK 5: 8,225 words, or 75.2 percent.

1,800 characters, through HSK 6: 9,223 words, or 84.3 percent.

The full character set, including the 7 to 9 band: all 10,943 words.

Look at the third line again. Nine hundred characters is 30 percent of the character list, and it unlocks more than half of the vocabulary. That is the entire leverage of Chinese in one number. Your first thousand characters are worth several times your second thousand, and your fifth thousand is worth almost nothing to a learner.

One honest caveat about that table, because it matters. "Unlocked" here means every character in the word is one you have met. It does not promise you will know what the word means. You can know 电 and 脑 perfectly well and still not guess that electric brain is a computer. The table measures the wall coming down, not the meaning walking in. But the wall is the hard part, and this is what taking it down looks like.

So what should you aim for?

Pick by what you want to do, not by a round number.

For travel, menus, signs and short messages, 300 to 600 is enough. This is HSK 1 and 2 territory and it arrives faster than you think, because the characters on signs are the frequent ones.

For graded readers, chatting with a patient friend and simple articles, aim at 900 to 1,200. That is more than half the standard word list, and the point where studying stops feeling theoretical. If you never get further than this, you have still got real value out of Chinese.

For novels and news with a dictionary at your elbow, 1,500 to 2,000. You will look things up constantly at first and progressively less. Most learners who reach here keep going, because the reading itself starts doing the teaching.

For comfortable, uninterrupted reading, 2,500 to 3,000. The full HSK 3.0 character set is 3,000 characters, so the top of the standard and the top of the useful range land in the same place.

Notice what is missing from that list: any number above 3,000. As a learner, you will essentially never have a reason to chase one.

Stop guessing and measure the thing you actually want to read

Everything above is an average, and you are not going to read an average. You are going to read one specific novel, one specific news site, or the messages one specific person sends you.

So measure that instead. Paste it into our Chinese text difficulty checker and it will tell you the HSK level of that text, your coverage of it at whatever level you are at now, and the exact list of characters standing between you and it. It runs entirely in your browser, the text is never uploaded, and it takes about five seconds.

That list of missing characters is almost always shorter than people expect. Wanting to read something specific, and then discovering it is forty characters away rather than a thousand, is the best motivation this language has to offer.

Which characters matters more than how many

One last thing, and it moves timelines more than any target you pick.

Two learners can both know 800 characters and be nowhere near each other in reading ability, because it is the composition of those 800 that counts. Characters learned in frequency order compound, because the frequent ones keep reappearing and reinforcing themselves for free. Characters learned at random do not.

There is a structural version of the same point. Characters are built from a small set of reusable parts, so learning 木, tree, before 林, grove, and 森, forest, makes the second and third nearly free. Meet them in the wrong order and each one is a fresh wall of strokes. I have written about how components actually work if you want the mechanics, and about what makes characters stick once you have learned them.

Which is a long way of saying: the honest answer is 1,500 for most people and 3,000 to be genuinely comfortable, but the number matters far less than the order you meet them in and whether they are still there next month.

That second part is the harder problem, and it is the one Hanzi Express is built around: components before the characters that use them, characters before the words they build, and spaced repetition scheduling reviews so the 900 you have learned are still 900 in March instead of 600. The first three levels are free, no card required, which is plenty to feel whether this approach fits the way you learn.

How Many Chinese Characters Do You Actually Need to Know?