The 100 most frequent characters in everyday Chinese sentences, in order: pinyin, meaning, each one’s share, and how much you cover once you have learned up to it.
Chinese Characters by Learning Order (汉字学习顺序)
How much of these sentences you can read at each milestone
Each bar is the share of characters in these sentences that you would recognise after learning every character up to that rank. On the right is the character that sits at that rank.
- Top 1020.7%有
- Top 10053.2%行
- Top 50083.1%万
- Top 1,00093.2%禁
- Top 2,00098.8%鸣
The yardstick schools use
- End of Year 6
- About 3,000 common characters recognised, about 2,500 of them written
- End of Year 9
- About 3,500 recognised
Data: RECATOOLS’ own count of 88,085 Mandarin sentences from Tatoeba (export of 31 October 2026), 895,075 characters, CC BY 2.0 FR, credit to the contributors of Tatoeba. School targets: China’s Compulsory Education Chinese Curriculum Standard (2022 edition), pp. 11 and 14. Tatoeba downloads · Curriculum PDF (Chinese) · Frequency in Chinese Wikipedia
Using the list
Type to filter
Type a character in either script, pinyin (tone marks optional — wo works), an English meaning or a rank, and the table narrows as you type. Clear the box to get the full list back.
Read the last column
“Covers so far” is the running total up to that character. By 学 at rank 44 you already recognise 39.9% of the characters in these sentences.
A signpost, not a syllabus
The top of the list is mostly particles and pronouns, which mean little on their own. Learn them inside common words and sentences.
Why the first hundred characters do so much work
Chinese characters are used very unevenly. In these 88,085 sentences, 我 alone is 3.9% of everything, the top 10 together are 20.7%, and the top 100 are 53.2%. After that each new character adds less: going from the 1,000th to the 2,000th means learning 1,000 more for 5.6 more percentage points.
Rank depends on what was counted. These sentences are mostly everyday speech, so 我 is 1st, 吗 21st and 吃 72nd. In our count of Chinese Wikipedia, 我 is only 465th and 年 is 2nd, against 75th here.
The sentences are short and plain, collected for language learners, so the curve is steeper than for ordinary text: the same top 100 make up only 35.8% of Chinese Wikipedia. The corpus is small too, about 900,000 characters. The character at rank 100 appears about 1,570 times, so the order of neighbouring ranks means nothing.
Recognising characters is not reading words. Knowing 93% of the characters means about one in fourteen is still unknown, and plenty of words mean something other than their two characters added together.
Questions about the list
RECATOOLS counted them from Tatoeba’s Mandarin sentences, export of 31 October 2026, 89,248 sentences. After converting to Simplified we dropped 1,163 duplicates and removed the stock names 汤姆 and 玛丽 (Tom and Mary), which would otherwise put 汤 and 姆 at ranks 18 and 19. That leaves 88,085 sentences and 895,075 characters, used under CC BY 2.0 FR with credit to Tatoeba’s contributors. Until November 2026 this page used Jun Da’s frequency list, whose copyright notice asks for written permission before commercial use, so we replaced it with our own count.
No. This is a count of characters, not a syllabus, and no HSK list was used to make it.
Not in that sense. 93.2% is the share of characters in these sentences you would recognise, not the share of meaning, and it would be lower for a newspaper or a novel. The new characters tend to be the content words that carry the information, and many words mean more than their characters do separately.
The count found 4,073 different characters, and the further down the list, the fewer times each appears and the less the rank means. The top 100 are the steadiest and the best return on effort. The chart above carries the curve out to rank 2,000.
Under the 2022 curriculum standard: about 3,000 common characters recognised by the end of Year 6, about 2,500 of which they can write, and about 3,500 recognised by the end of Year 9.
Method & sources
How it computes
RECATOOLS count of Tatoeba's Mandarin sentences, export of 2026-10-31 (89,248 sentences): OpenCC t2s, 1,163 duplicates across scripts dropped, the stock names 汤姆 and 玛丽 removed, CJK ideographs only (88,085 sentences, 895,075 ideographs, 4,073 distinct). Share = a character's count ÷ the total; 'Covers so far' = the running share up to that rank. Held in CharacterFrequencyKernel. CC BY 2.0 FR, credit to Tatoeba's contributors.
What this tool implements
- The list used is named on the page: Tatoeba's Mandarin sentences, export of 31 October 2026, counted by RECATOOLS. It is a count of characters, not a syllabus. No HSK list was used.
- Short sentences collected for language learners are not general text: the top 100 make up 53.2% here and 35.8% of Chinese Wikipedia. The page says so, and says the order of neighbouring ranks means nothing at this size (rank 100 appears about 1,570 times).
- Corrected 2026-11-03: the table, percentages, cumulative column and chart were Jun Da's Modern Chinese list, whose notice asks for written permission before commercial use; replaced by our own count. Tone-free pinyin search, which the page promised, did not work (only toneless entries such as "de" matched); it does now. A Traditional search found nothing; it does now, and zh-TW leads with Traditional.
- School targets re-read from the curriculum PDF 2026-11-03: p. 11 and p. 14 say what the page says. The line tying 3,500 characters to 99.5% of text was Jun Da's figure and is gone.
Sources
- Tatoeba. Per-language export, Mandarin sentences (cmn_sentences.tsv.bz2), modified 2026-10-31; sentences CC BY 2.0 FR. https://tatoeba.org/en/downloads
- Tatoeba. Terms of use, section 6.2, Creative Commons licenses applicable to sentences. https://tatoeba.org/en/terms_of_use
- Ministry of Education (PRC). 义务教育语文课程标准（2022年版）, p. 11, 第三学段: 「累计认识常用汉字3000个左右，其中2500个左右会写」; p. 14, 第四学段: 「累计认识常用汉字3500个左右」. http://www.moe.gov.cn/srcsite/A26/s8001/202204/W02022042058234438…
- OpenCC, the conversion tables used for t2s and for the Traditional forms shown (Apache-2.0). https://github.com/BYVoid/OpenCC
Pick up where you left off
Stored only in this browser — never sent to our servers.