Free Chinese reading level test

How Many Chinese Characters Do You Know?

Rate 40–70 characters, pass a few quick spot-checks, and get an honest estimate — plus your HSK level and the stories you can read today. About 3 minutes, in Simplified or Traditional.

What makes this test different

• It checks, not just asks. Most online character tests are pure self-rating, which rewards optimism. Here a share of your “I know it” answers are verified, and the estimate is corrected by how those checks went.
• One script at a time. Simplified and Traditional are separate modes, so a Taiwan learner is never shown 说 and a Mainland learner is never shown 說.
• It adapts. Instead of a fixed list, each character is picked from the 3,059-character ranking where your answer tells us the most, so beginners aren’t buried in rare characters and advanced readers aren’t bored by 的 and 是.
• The number comes with meaning. You get a likely range, your approximate HSK 3.0 level, how much of real story text you can cover, the free stories you can read today, and a list of the characters you missed.

What your number means

Characters are not equally useful. The table shows what knowing the top N characters of our ranking buys you, measured on 702 stories from the Literate Chinese app library (148,466 character tokens, names excluded) and the 277 stories you can read free on this site. A story counts as readable when you know at least 98% of its characters — the coverage level reading research links to comfortable, unassisted reading.

Two things stand out. Coverage climbs fast — 87% of story text with just 500 characters — but whole stories need more, because a story is only comfortable when almost every character is familiar. And the jump between 500 and 1,500 characters is where graded reading really opens up: 24 readable web stories become 183.

How many characters do you need?

For exams, the 2021 HSK 3.0 standard lists its characters level by level — 300 new characters at each of levels 1–6 and about 1,200 for HSK 7–9. These are the counts in this test’s pool (a handful of characters used only in names or as variant forms are left out):

Beyond exams, the commonly cited figure for newspapers and native novels is 2,500–3,000 characters; China’s 1988 list of 3,500 common characters covers about 99.5% of general published text, and an educated native reader recognises roughly 3,000–4,000. We walk through the whole ladder, with our corpus data, in How Many Characters Do You Need to Read Chinese?, and put the big numbers in context in How Many Chinese Characters Are There? Note that those articles rank characters by frequency in our own catalog, an idealised learner who knows exactly the catalog’s most common characters, so their story counts run a little higher than the ones above.

How the estimate works

1. A ranked pool of 3,059 characters. Every character in the HSK 3.0 syllabus and Taiwan’s TOCFL word lists, ordered by a blend of four signals: the HSK level that introduces it, the TOCFL level that introduces it, how often it appears in our 702-story corpus (character names excluded), and how many dictionary words use it. Vulgar characters and characters used only in names are left out, and the Simplified pool contains only Simplified forms.
2. Adaptive sampling. The first few characters move up or down the ranking quickly to find your level; after that, each character is chosen where you are most likely to be unsure, with an occasional character from anywhere on the scale to catch gaps. Most people rate 45–70 characters.
3. A curve, not a cutoff. Nobody knows exactly the top N characters and nothing else. The test fits a smooth curve — the chance you know a character, falling with its rank — and adds up that curve across all 3,059 characters. The likely range is the middle 80% of plausible curves.
4. Spot-checks against over-claiming. Self-rated tests tend to run high, mostly because “I’ve seen that” slips into “I know that”. A share of your “I know it” answers — more of them near your limit — get a four-option meaning or pinyin question with wrong answers drawn from characters of similar frequency. Checked answers count as they turned out, and unchecked ones are weighted by your measured accuracy, corrected for lucky guesses.

Honest limits. We tested the engine on 270 simulated test-takers, from near-beginners to advanced readers. Answering honestly, their median error was 9%; for those who knew 600+ characters about 83% of estimates fell within ±15% of the true count (70% across all levels — small numbers swing more in percentage terms), and the true count sat inside the stated range 80% of the time. With heavy over-claiming (a 30% chance of claiming characters they didn’t know), spot-checks cut the median error from 38% to 13%. The scale tops out at 3,059 characters, so near-native readers will hit the ceiling. Characters are not words: knowing 电 and 脑 does not guarantee you know 电脑 (computer). And the ranking reflects learner materials and exam lists more than newspapers, so a reader whose Chinese comes mainly from news or literature may score a little differently.

We considered adding made-up “fake characters” as a lie detector, as some English vocabulary tests do with fake words. We left them out: drawing convincing fake characters takes images rather than text, and asking you about the meaning of a real character is a fairer check than trying to trick you.

Frequently asked questions