The Chinese languages, by number of speakers

Chinese is not one language with accents. It is a family with eight or ten branches, and the fifth-largest of them has about forty-eight million speakers and almost no presence in English at all.

Updated 22 September 2026 · 4 min read

The branches

Wikipedia’s survey of the varieties of Chinese gives the share of first-language Chinese speakers each top-level branch has. Beside it, the best available absolute figure, which comes from the individual branch articles and mostly from Ethnologue:

Taken together, about 1.39 billion people — some seventeen per cent of everyone alive — speak a variety of Chinese as their first language.

Read those numbers with care

This is the part most listicles skip, and it is the reason two sites will give you different figures for the same language.

• The percentages and the absolute numbers come from different places and different years. That is why Wu takes a larger share than Yue in the percentage column while Yue has more speakers in the absolute column. Both can be sourced; they are not measuring the same year.
• Gan’s own Wikipedia article contradicts itself, giving 23 million in its summary and a province-by-province total of about 48 million in its text. I have printed both rather than pick one.
• Mandarin’s share is given as 65.7% in one Wikipedia article and 70% in another. Same encyclopaedia, different pages.
• Branch membership is itself disputed. Jin was inside Mandarin until the linguist Li Rong proposed in 1985 that it be treated as its own top-level branch. Huizhou shares features with Wu, Gan and Mandarin and is, in Wikipedia’s words, “difficult to classify”.

None of this makes the table useless. It means the right way to read it is as orders of magnitude: Mandarin is enormous, four branches are in the tens of millions, and the rest are smaller but not small.

The one nobody has heard of

Jin. Around forty-eight million speakers, which is more than Hakka and more than Xiang, spoken across Shanxi and into Inner Mongolia — and it is the only major non-Mandarin branch spoken predominantly in northern China. It was separated out from Mandarin in 1985 partly because it kept the entering tone as a glottal stop, which its neighbours lost.

Xiang, in Hunan, has thirty-eight million. Gan, in Jiangxi, has somewhere between twenty-three and forty-eight. Each is comparable in size to a well-known European national language, and each has almost no English-language presence — no courses, very little written about them, nothing a learner could pick up.

The asymmetry is worth noticing. Cantonese, at eighty-five million, is a world language with dictionaries, films and a written form. Jin, at forty-eight million, is a Wikipedia article. The difference is not size; it is Hong Kong, emigration and a film industry.

Standard Chinese is not the same thing as Mandarin

This trips up nearly everyone, including people who speak it. Mandarin is a branch of about 990 million people stretching across most of China, internally varied enough that its ends are hard going for each other. Standard Chinese — Putonghua — is a single standardised register: its sounds from the Beijing dialect, its vocabulary from the Mandarin group, and its grammar from modern written vernacular literature.

So when a figure says eighty per cent of China speaks Mandarin, it means the standard, learned at school, and it is a different claim from saying eighty per cent natively speak a Mandarin variety. China reported passing that eighty per cent mark in 2020, up from about fifty-three per cent in 2007 — an extraordinarily fast change in one generation, and the background to the question of whether the other varieties are dying.

Are these languages or dialects?

The branches are not mutually intelligible with one another, and several are not internally intelligible either — Min is the extreme case, where Wikipedia notes that varieties in neighbouring counties, and in the mountains of western Fujian even adjacent villages, are often mutually unintelligible.

Whether that makes them separate languages is a question about the word rather than about the facts, and it has a page: what a Chinese topolect is. The short version is that ISO 639-3 and Ethnologue assign each branch its own language code, while the official position of the PRC is that they are variants of one Chinese language.

Where to go from here

• Wu, the branch Shanghainese belongs to, is on the Wu Chinese page — its groups, its internal divisions, and why northern and southern Wu cannot understand each other.
• Yue set beside Wu is on the Shanghainese and Cantonese page: two survivals of Middle Chinese that took different routes.
• Mandarin against Shanghainese, feature by feature, is on the comparison page.

And if you want to hear one of them rather than read about it, the Shanghainese course is free and has recordings of every word it teaches.

Sources

1. Wikipedia: Varieties of Chinese — the branch shares, the classification caveats, and the ISO 639-3 and PRC positions
2. Wikipedia: Mandarin Chinese — 990 million first-language speakers
3. Wikipedia: Wu Chinese — 83 million
4. Wikipedia: Yue Chinese — 85 million
5. Wikipedia: Min Chinese — 75 million, and the internal unintelligibility
6. Wikipedia: Jin Chinese — 48 million; Li Rong’s 1985 proposal; the entering tone as a glottal stop
7. Wikipedia: Gan Chinese — the 23 million summary figure and the provincial total of about 48 million
8. Wikipedia: Hakka Chinese — 44 million
9. Wikipedia: Xiang Chinese — 38 million
10. Wikipedia: Chinese language — about 1.39 billion first-language speakers, 17% of the world
11. Wikipedia: Standard Chinese — the standardised register, and the reported 80% figure for 2020 against 53.06% in 2007