Chinese name gender
Han characters and pinyin both work. Coverage is near total and accuracy is the lowest of any major market — that inversion is a property of how Chinese given names are formed, and this page says so rather than averaging it away.
100 free credits every day. No card.
Those four figures are one country. Measured across the whole writing system instead, wherever the name is used, the same fixture gives 222 names, 100.0% coverage, 90.5% when answered and 90.5% end-to-end. If your data is not from a single country, the wider figure is the one to plan against.
Measured 7 October 2026 on a fixed, reproducible fixture. Not a population sample — a fixed input anyone can rerun against us or against a competitor.
Twenty names in China’s 2021 birth registrations
The Ministry of Public Security lists these exact two-character given names among its ten most frequent names for each newborn cohort. The list is a year-specific example, not a probability for a person with that name. Its report gives approximately 23,000 boys named 沐宸 and 18,000 girls named 若汐; their measured profiles are linked below.
| Boys’ top ten | Girls’ top ten |
|---|---|
| 沐宸 | 若汐 |
| 浩宇 | 一诺 |
| 沐辰 | 艺涵 |
| 茗泽 | 依诺 |
| 奕辰 | 梓涵 |
| 宇泽 | 苡沫 |
| 浩然 | 雨桐 |
| 奕泽 | 欣怡 |
| 宇轩 | 语桐 |
| 沐阳 | 语汐 |
Source: 2021 National Name Report (Ministry of Public Security)
Why accuracy is lower here than in Europe
European given names come from a finite inherited stock: a name carries its gender as membership in a list. Chinese given names are composed — a parent picks one or two characters for meaning, sound and balance — so the pool is effectively open and the signal lives in connotation rather than membership.
Many characters are genuinely used for both. 静 (jìng, "quiet") leans female, but only at 68. Reading the gender field without the probability turns a real 68 into a fake certainty.
Real output, country hint CN
These API results return sample_size: 0 and confidence: unverified because the current lookup source supplies no per-name observation counts. A separate official 2021 report publishes selected newborn name counts, shown above; those counts are not yet part of the API lookup.
| Input | Gender | Probability |
|---|---|---|
| 伟 (Wei) | male | 85 |
| 静 (Jing) | female | 68 |
| 芳 (Fang) | female | 85 |
| 强 (Qiang) | male | 87 |
| 李伟 (Li Wei) | male | 92 |
| 王芳 (Wang Fang) | female | 93 |
Send characters rather than pinyin
Romanisation loses information twice. Tone marks are dropped, so wei collapses 伟, 薇, 微, 威 and 唯 into one string. Character identity is dropped even at the same tone. Pinyin is a lossy hash of the thing you actually wanted to look up, and the answer you get back is the blend.
If your source system stores characters, send characters. If it stores pinyin, send pinyin and lower your threshold.
The surname-first trap
Chinese names are written surname first, and romanised lists use both orders. Until 9 October 2026 a two-word pinyin name was read first word first, so Li Zhiyuan classified Li — a surname, which carries a weak population-level gender lean. On 174 Chinese names written that way accuracy was 49%, mostly at high probability.
Now, when one of two pinyin words is a common Chinese surname, that word is taken as the surname on either side; when both could be, the more common one is. Zhang Wei and Wei Zhang both classify Wei, and both orders score 87.4% on the same names. It misses when the given name is a more common surname (Liu Li) and does not cover Cantonese spellings, so put the given name in the name field where you can and read the name field back to see which token won.
Single characters are whole names
A one-letter token in J. Smith is an initial and should be skipped; a single Han character is a complete given name. A Latin-centric normaliser drops exactly the names this market is made of, which is a bug you will not see unless your test fixtures are non-Latin.
Questions
Do you support traditional characters?
Both simplified and traditional inputs reach the lookup; the entry matched is reported in the name field.
Should I pass country=CN?
Pass it when you know it. It narrows the pool. The current CN lookup source has no observation counts, so its answers carry confidence unverified even though a separate official report publishes counts for selected names.
Why is my Chinese answer unverified but correct?
Because confidence reports counted evidence in the lookup, not correctness. The current CN lookup source has no counts, so it returns unverified even for names that appear in a separate official report.
What about Hong Kong, Taiwan and Singapore?
Names resolve the same way; pass the relevant country code when you have it, and expect romanisation conventions to differ from mainland pinyin.
Related pages
Predict gender from a Japanese name. Our measured accuracy here is the lowest of any market we publish, the sample is small, and this page shows both.
Predict gender from a Korean name in Hangul or romanised form. Measured coverage and accuracy, competing romanisation systems, and surname-first handling.
Every name lookup returns a probability, a sample size and a confidence tier. What each one measures, how the tiers are set, and where to put your threshold.
Coverage on Han names is complete but accuracy is 90.5% — the inverse of the usual pattern. Why pinyin is weaker evidence, and the name-order trap.
A reproducible benchmark across 4,610 names, 25 countries and nine writing systems. Thai still answers only half the names, and we publish those numbers too.
Check it against your own list
Every number on this page is reproducible with a free key. If your data breaks it, that is the more interesting result.