Chinese proficiency test HSK7-9 paper 4, HSK 3.0 format

91 questions 30 + 60 + 55 + 35 min

Which format is this?

This test follows the renewed HSK 3.0 format, published by Chinese Testing International in December 2025 and due to apply from July 2026.

Sittings for levels 1 to 6 still use the previous format at the time of writing, and no changeover date has been announced. If your exam is soon, expect the older paper on the day.

The vocabulary is the same either way: the HSK 3.0 word list is what you have to know for both formats, and that is what this test drills.

Listening

Each recording is played once only, as in the real exam: the button will not play it again.

Part 1 (1–10)

Listen, then say for each sentence whether it matches what you heard. Each sentence is judged on its own.

  1. 1

    ★ 这些旧楼建成的时候,按当时的规定不必装电梯。

  2. 2

    ★ 新办法规定,住户按楼层高低分担费用,越高出得越多。

  3. 3

    ★ 真正卡住这件事的主要是技术上的困难。

  4. 4

    ★ 从提出申请到装好,全部过程大约需要两个月。

  5. 5

    ★ 有居民担心加装电梯以后楼道会变窄。

  6. 6

    ★ 这片古代的村落到今天大约有五百年。

  7. 7

    ★ 按照规定,挖出这样的东西以后工程要立即停下。

  8. 8

    ★ 本市决定把这片古代的村落全部原地保留。

  9. 9

    ★ 工程推迟了五个月,多出来的费用由市里承担。

  10. 10

    ★ 这一类发现在城市建设当中非常少见。

Part 2 (11–22)

Listen, then either choose the right answer or fill in the gap, depending on what the question asks. A Chinese keyboard is required for the gaps.

  1. 11

    ★ 从家门口到公交站,中间只要有一级( ),前面做得再好也等于零。

  2. 12

    ★ 男的认为“只按轮椅设计”的问题是什么?

  3. 13

    ★ 下面哪一类人不在男的提到的需要无障碍的人当中?

  4. 14

    ★ 坡道修好了,尽头却放一根( ),这是常见的断点之一。

  5. 15

    ★ 关于那两种常见的断点,男的怎么看?

  6. 16

    ★ 听不见的人需要什么样的办法?

  7. 17

    ★ 那所学校走廊的扶手有什么特别的地方?

  8. 18

    ★ 男的认为好的无障碍设计通常有什么特点?

  9. 19

    ★ 在设计阶段就加入无障碍,成本大约会增加多少?

  10. 20

    ★ 房子盖好以后再改,费用会怎么样?

  11. 21

    ★ 男的为什么说“别把它当成对别人的照顾”?

  12. 22

    ★ 这段访谈主要谈的是什么?

Part 3 (23–40)

Listen, then either choose the right answer or fill in the gap, depending on what the question asks. A Chinese keyboard is required for the gaps.

  1. 23

    ★ 上个世纪五十年代到八十年代,这座城市的社会结构有什么特点?

  2. 24

    ★ 变化为什么从九十年代开始?

  3. 25

    ★ 第一次的做法为什么不成功?

  4. 26

    ★ 人们后来总结说,那时候并没有真正换掉( )。

  5. 27

    ★ 第二次的做法包括哪些?

  6. 28

    ★ 最早反对建这处场馆的是谁?为什么?

  7. 29

    ★ 古代的园林占地不大,为什么走进去却觉得走不完?

  8. 30

    ★ 进门迎面放一堵墙或者一块石头,作用是什么?

  9. 31

    ★ 把园子外面的树和塔借进画面,这种做法叫什么?

  10. 32

    ★ 地上的路面故意做得不平,是为了什么?

  11. 33

    ★ 说话人认为今天很多新建园子的问题是什么?

  12. 34

    ★ 说话人认为园林真正精巧的地方在哪里?

  13. 35

    ★ 耳朵里那个很小的器官负责感觉方向和( )。

  14. 36

    ★ 坐车看书的时候,眼睛报告的是什么?

  15. 37

    ★ 身体为什么会难受?

  16. 38

    ★ 在漫长的过去,让两套信息对不上的最常见原因是什么?

  17. 39

    ★ 录音建议坐在前排,理由是什么?

  18. 40

    ★ 关于“习惯就好了”这个说法,录音怎么看?

Reading

Part 1 (41–68)

Read, then answer the question.

  1. 41

    “分期不收利息”这几个字很容易让人以为借钱不要钱。实际上,钱和别的东西一样,用别人的就要付价格,这个价格就是利息。免掉的只是名字,不是数目。最常见的做法有两种。第一种是商家先把利息算进价格里:同一件东西,一次付清比分期便宜,差出来的那一部分就是。第二种是商家替你把这笔钱交给了借钱给你的那一方,作为交换,它拿到了这一笔生意——对它来说这是广告的费用,对你来说仍然是价格,只是不写在你的单子上。真正需要看清楚的是利率,而利率最容易被误会。一个常见的说法是“每月只收千分之五”,听上去很低,一年不过百分之六。可是这个说法把你还回去的钱当成一直没有还,而实际上你每个月都在还,手上剩下的本来就越来越少。把这一点算进去,真实的利率大约是那个数字的( )。这不是谁存心骗人,是两种算法的区别,可是差得很多。第三件事是时间。同样一笔钱,分十二个月和分二十四个月,每月的数目差不多能减一半,很多人因此选长的。代价是总共付出去的钱明显更多,而且在那两年里,你的选择被绑住了。所以判断的顺序应该是:先问这件东西是不是现在必须买,再问一次付清是多少,最后才比较分期的条件。把顺序倒过来的人,多数会买下本来不打算买的东西。

    ★ 作者认为“分期不收利息”真正免掉的是什么?

  2. 42

    ★ 商家把利息算进价格里,怎么才看得出来?

  3. 43

    ★ “每月只收千分之五”这个说法为什么容易误会?

  4. 44

    ★ 按照上下文,空白处应该填的是:

  5. 45

    ★ 分二十四个月比分十二个月的代价是什么?

  6. 46

    ★ 作者建议的判断顺序是什么?

  7. 47

    ★ 上面这篇文章想说明的是:

  1. 48

    中国城市里的住宅小区大多有一圈墙,门口有人看着,进出要刷卡。这在今天看来理所当然,其实历史并不长。把整片住宅围起来的做法,是上个世纪后期随着大规模建设一起出现的。它解决了当时最急的两件事:一是安全,二是管理——有了边界,谁负责这一片就清楚了。代价当时并不明显,如今却越来越重。第一重代价是路。一个占地二十公顷的小区,等于在城市的路网上挖了一个洞。旁边的人要绕一公里,公交车开不进去,路口因此更堵。城市看上去有很多路,可是能走的路被墙切断了。第二重代价是街。临街的一层如果朝里开,那么对着马路的就只是一堵墙。一条两边都是墙的路,人走在上面既不安全也不愿意停留;而一条两边有小店的路,哪怕更窄,人反而更多。街道的活力不是靠宽度,是靠( )。所以近年出现了“打开小区”的说法。反对的声音很实在:安全怎么办?停车怎么办?谁来打扫?这些问题不能靠一句口号回答。实际做得比较好的地方,用的都是折中的办法——不是把墙全部推倒,而是在最需要的地方开几个口子,把内部的一两条路交给公众,同时把单元门那一层的安全做扎实。这件事的道理其实很朴素:安全应该做在门上,而不是做在整片土地的边界上。

    ★ 把整片住宅围起来的做法是什么时候出现的?

  2. 49

    ★ 这种做法当时解决了哪两件事?

  3. 50

    ★ 作者说一个大小区“等于在路网上挖了一个洞”,指的是什么?

  4. 51

    ★ 按照上下文,空白处应该填的是:

  5. 52

    ★ 反对“打开小区”的实在理由包括什么?

  6. 53

    ★ 实际做得比较好的地方用的是什么办法?

  7. 54

    ★ 作者最后想说明的道理是什么?

  1. 55

    过去二十年,中国的乡村小学数量减少了一大半。原因并不复杂:出生的孩子少了,年轻的父母进城了,一个村子里到了上学年龄的孩子从几十个变成几个。把他们集中到镇上的一所学校去,看上去是唯一合理的做法。合并确实带来了实在的好处。镇上的学校有专门教外语和音乐的老师,有实验室,有操场;而一所只剩下七个孩子的村里小学,一个老师要教全部的课,从语文教到体育。从这个角度看,合并提高的是每一个孩子能得到的东西。代价出现在路上。孩子每天要走的路从五分钟变成四十分钟,冬天天不亮就要出门。为了解决这个问题,多数地方办了住的地方,让孩子一周回一次家。于是一个六岁的孩子开始在集体的房间里生活,而他的父母多数不在身边。研究这件事的人反复提到一个现象:合并以后成绩确实提高了,可是低年级孩子的( )明显变差,尤其是最小的那两年。另一个后果落在村子上。学校常常是一个村子里唯一每天都有人聚在一起的地方,读书声一停,村子就更空了。有的地方后来发现,学校一撤,最后几户还在犹豫的年轻人也搬走了。近几年做法有所调整:低年级尽量留在村里,高年级再合并;同时用车接送代替住在学校。这不是回到从前,而是承认一件事——教育的好坏不只写在成绩单上,六岁的孩子每天晚上睡在哪里,也是教育的一部分。

    ★ 乡村小学数量减少的原因是什么?

  2. 56

    ★ 合并带来的好处是什么?

  3. 57

    ★ 一所只剩下七个孩子的村里小学是什么情况?

  4. 58

    ★ 按照上下文,空白处应该填的是:

  5. 59

    ★ 学校撤走以后,村子里发生了什么?

  6. 60

    ★ 近几年的调整是什么?

  7. 61

    ★ 作者最后想说明的是什么?

  1. 62

    你在手机上按下购买,两天以后东西到了门口。这两天里发生的事,比多数人想的复杂得多,而它有一个名字,叫供应链。供应链的起点不在工厂,在预测。工厂要在你下单之前几个月就决定生产多少,因为原料的采购、机器的安排都需要时间。预测错了,代价立刻出现:多了,仓库里堆着卖不掉的东西,占着钱;少了,你看到的是“无货”,而这个损失不写在任何一张单子上。第二段是仓库的位置。很多人以为东西是从工厂直接寄给你的,实际上它多半在你下单以前就已经躺在离你不到三百公里的一个仓库里。平台根据过去的销售,把不同的商品提前分到不同的城市——真正让“两天到货”成为可能的是这一步,而不是运得更快。第三段才是运输,而这一段里最花时间的不是长途,是最后那几公里。一辆车从这座城市开到那座城市要八个小时,从分货的地方送到你家门口的四公里,可能要花掉半天,因为它要停几十次。整条供应链上最容易断的地方,往往是( )的一环。一场大雨、一次系统出错、一个小零件的短缺,都会让后面全部停下。这些年反复出现的教训是:一条为了最省钱而设计的供应链,通常也是最容易断的一条。省下来的是每天的钱,付出去的是断掉那几天的钱。

    ★ 供应链的起点是什么?

  2. 63

    ★ 预测少了会有什么后果?

  3. 64

    ★ 真正让“两天到货”成为可能的主要是什么?

  4. 65

    ★ 运输这一段里最花时间的是哪一部分?

  5. 66

    ★ 最后那几公里为什么慢?

  6. 67

    ★ 按照上下文,空白处应该填的是:

  7. 68

    ★ 这些年反复出现的教训是什么?

Part 2 (69–73)

Put the paragraphs back in order. Watch out: one paragraph fits nowhere, and another is already in place.

  1. A 此后三十年,他管的这一段线路走过七座山。哪一根杆下面的土是松的,哪一段容易被树枝碰到,他闭着眼也说得出来。他给自己定的那条规矩后来成了整个班组的规矩,写在墙上。
  2. B 第一次上杆是在训练的场地,离地八米。他后来说,难的不是高,是手上要同时做三件事,而脚下只有两个铁钉。带他的师傅在下面只说了一句:慢一点,你有的是时间。
  3. C 他家在山里,小时候村子每天只来两个小时的电,天一黑就得点灯。十九岁那年,供电的单位到村里招人,招的是能爬杆的。全村报名的有二十多个,最后留下三个,他是其中之一。
  4. D 参加工作的第三年冬天,一场大风把山那边的线刮断了,两个村子停了电。他们连夜上山,风里能看见的不到十米。那一次他在杆上待了四个小时,下来的时候手已经握不住东西。事情办成了,可是从那以后,他给自己定了一条规矩:再急也先看天。
  5. E 这一带的山里长着大片的槐树,每年五月开花,白色的一串一串,风一吹满山都是味道。附近的人常常在这个时候上山来,采一些回去做菜。
  6. F 最近十年,工具变了。检查线路不再靠人一根一根爬上去看,改成用无人的小飞机飞一趟;有些偏远的段落,重的东西也由直升的飞机吊上去。他一开始不太服气,用了两年以后承认,机器看见的东西比人多,而人省下来的力气可以用在真正要爬的那几次。
  7. G 如今他快六十了,主要在下面带人。有人问他这些年最怕的是什么,他说不是高,也不是风,是一个新人在杆上忽然显得很有把握的样子。他说,那种时候你叫他,他往往听不见。

D 3rd position · already placed

  1. 69 1st position
  2. 70 2nd position
  3. 71 4th position
  4. 72 5th position
  5. 73 6th position

Part 3 (74–87)

Read the passage, then answer each question in ten Chinese characters or fewer. A Chinese keyboard is required.

Practice — this part is not marked

The real exam asks you to write freely here, and a human examiner reads it. No browser can mark that, and this page does not invent a score: these questions do not count towards your result. Write them anyway — it is the only practice this part allows, and a possible answer is offered once you have handed in.

  1. 74

    夏天的夜里,房间里有二十个人,被咬的常常总是同一个。这不是运气,蚊子确实在挑。它找人分三步,距离由远到近,用的信息完全不同。最远的一步靠人呼出的气。人呼出的气里有一种它认得的成分,在几十米之外形成一条断断续续的带子,蚊子逆着这条带子飞过来。所以在户外,说话多、动得多、呼吸重的人先被找到,孩子和睡着的人反而安全一些。第二步在几米之内,靠身上的味道。皮肤表面住着大量很小的生物,它们把汗变成各种气味,而每个人的组合都不一样。研究者做过一个试验:让蚊子在两只袜子之间选,它们几乎每次都选同一只。这一步解释了为什么被咬的总是那几个人——那是天生的,洗澡也改不了多少。第三步在几十厘米之内,靠温度和湿度。这也是为什么它总是盯着脚和手:那里离衣服远,热和湿的信号最清楚。知道了这三步,办法就清楚了。赶蚊子的东西起作用不是因为它难闻,而是因为它盖住了第二步需要的信息,让蚊子在最后几米找不到你。风扇有效也不是因为风把蚊子吹走了——它飞得动——而是因为风把那条气的带子吹散了。至于灯,蚊子对它并不感兴趣,那是被别的虫子连累的误会。

    ★ 蚊子找人一共分几步?

    0 Chinese characters written

    Show a possible answer

    三步

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

  2. 75

    ★ 最远的那一步靠什么?

    0 Chinese characters written

    Show a possible answer

    人呼出的气

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

  3. 76

    ★ 在户外,哪一类人先被蚊子找到?

    0 Chinese characters written

    Show a possible answer

    呼吸重的人

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

  4. 77

    ★ 第二步靠的是什么信息?

    0 Chinese characters written

    Show a possible answer

    身上的味道

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

  5. 78

    ★ 蚊子为什么总是盯着脚和手?

    0 Chinese characters written

    Show a possible answer

    那里热和湿最清楚

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

  6. 79

    ★ 赶蚊子的东西为什么起作用?

    0 Chinese characters written

    Show a possible answer

    它盖住了味道

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

  7. 80

    ★ 风扇有效的真正原因是什么?

    0 Chinese characters written

    Show a possible answer

    把那条带子吹散了

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

  1. 81

    很多人拿到一种药,第一件事是看说明上写的副作用,看完就不敢吃了。这个反应可以理解,可是它建立在一个误会上:写得越多,不等于越危险。副作用的意思是,药在做它该做的事情之外,还做了别的。原因在于药到了身体里不会只去一个地方,它跟着血走遍全身,而它要影响的那种反应,往往在别处也存在。所以副作用不是失误,是同一种作用出现在了不该出现的地方。说明上那一长串,来自试验和上市以后的记录,只要有人报告过就要写上去,哪怕只有万分之一,哪怕并不能确定和这种药有关。写得多,多数时候说明这种药被研究得久、用得广。真正需要注意的是排在前面并且写出了比例的那几条。第二个常见的误会和剂量有关。同一种药,少了没有用,多了变成毒,中间那一段才是有效的范围。很多人以为多吃能好得快,实际上是把自己推到了范围以外。老年人和孩子的这个范围比成年人窄,这也是他们的用量要单独算的原因。最后一点最要紧:不同的人、不同的病,用的药不同。同样是咳嗽,原因可以有十几种,邻居吃了有效的那一种,到你身上可能一点用也没有,甚至有害。这件事有一个流传很久的四个字的说法,讲的正是这个道理——( )。

    ★ 副作用的意思是什么?

    0 Chinese characters written

    Show a possible answer

    药还做了别的事

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

  2. 82

    ★ 药为什么会有副作用?

    0 Chinese characters written

    Show a possible answer

    它跟着血走遍全身

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

  3. 83

    ★ 说明上写得多,多数时候说明什么?

    0 Chinese characters written

    Show a possible answer

    被研究得久用得广

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

  4. 84

    ★ 那一长串里真正需要注意的是哪几条?

    0 Chinese characters written

    Show a possible answer

    排在前面写出比例的

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

  5. 85

    ★ 关于剂量,正确的说法是什么?

    0 Chinese characters written

    Show a possible answer

    少了没有用多了变毒

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

  6. 86

    ★ 老年人和孩子的用量为什么要单独算?

    0 Chinese characters written

    Show a possible answer

    他们的范围更窄

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

  7. 87

    ★ 上面这段文字的最后少了一个四个字的说法。请把它补上。

    0 Chinese characters written

    Show a possible answer

    对症下药

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

Writing

Part 1 (88)

Describe the chart, then analyse it, respecting the length asked for. A Chinese keyboard is required.

Practice — this part is not marked

The real exam asks you to write freely here, and a human examiner reads it. No browser can mark that, and this page does not invent a score: these questions do not count towards your result. Write them anyway — it is the only practice this part allows, and a possible answer is offered once you have handed in.

  1. 88

    请根据下面的图表,说明它反映的变化并加以分析,写一篇两百字左右的文章。

    Using the chart below, set out the change it shows and analyse it, in a continuous piece of about 200 characters. The exam allows 15 minutes for this question.

    某市城区和乡村中小学的数量
    The figures behind the chart
    2003 2008 2013 2018 2023
    乡村的学校 412 296 178 121 96
    城区的学校 58 84 121 163 198

    0 / 200 Chinese characters written

    Show a possible answer

    这张图表反映的是某市二十年里中小学数量的变化,每五年一个点。乡村的学校一路减少,从四百一十二所降到九十六所,只剩下最初的四分之一不到;城区的学校则一路增加,从五十八所升到一百九十八所,翻了将近三倍半。两条线在二零一三年前后交叉,此后差距继续拉大。把两条线放在一起看,能看出这不是一次简单的减少,而是一次搬家:孩子并没有消失,他们换了上学的地方。总数从四百七十所降到二百九十四所,减少的部分才是真正少掉的学生。这一点提醒我们,只看乡村那一条线,容易把两件事说成一件。当然,一张图也有它说不了的地方。它记录的是学校的数量,没有记录每一所学校有多少学生,也没有说明孩子上学的路因此变长了多少。

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

Part 2 (89)

Take a position on the question and argue for it, respecting the minimum length. A Chinese keyboard is required.

Practice — this part is not marked

The real exam asks you to write freely here, and a human examiner reads it. No browser can mark that, and this page does not invent a score: these questions do not count towards your result. Write them anyway — it is the only practice this part allows, and a possible answer is offered once you have handed in.

  1. 89

    “半途而废”是一个流传很久的说法,字面上的意思是:一件事做到一半就停下来,等于白做。有人把它当作对自己的提醒,认为多数人的失败不是因为选错了方向,而是因为停得太早;也有人认为,这个说法容易让人把已经花掉的时间当成必须继续的理由,心里清楚走错了路也不肯回头。请写一篇六百字左右的文章,谈谈你怎么看这个说法,并说明你的观点。

    A long-standing saying holds that stopping halfway is the same as never having started. Some take it as a reminder to themselves: failure rarely comes from picking the wrong direction, it comes from stopping too soon; others object that it turns time already spent into a reason to carry on, so that people persist down a road they know to be wrong. Write about 600 characters: how you read the saying, and where you stand.

    What the task asks for

    1. Open on the case where the saying is wrong, not the one where it is right: that is where the reader is waiting for you.
    2. Name what the saying leaves out — the time already spent, which comes back in neither case.
    3. Give two or three reasons, one per paragraph, each held up by a concrete case rather than a generality.
    4. Say how one tells persistence from stubbornness in practice: a position without a test is no use to anyone.
    5. Close on what your reading changes in the decision, not on a summary of what came before.

    0 / 600 Chinese characters written

    Show a possible answer

    我的看法是:这个说法给了一个正确的提醒,可是它用错了理由。停得太早确实是多数人失败的原因,但“已经做了一半”从来不是继续下去的理由。先说它为什么容易被用错。一件事做到一半,你已经花掉的时间、精力和钱,无论你接下来是继续还是停下,都拿不回来。既然两条路上它都一样,它就不应该出现在比较里。可是人偏偏最看重它,越是花得多越舍不得走,于是留在原地的理由变成了“我已经留了很久”。这不是坚持,这是被自己付出的东西绑住。第二,我承认反对的一半也有它的道理,而且这个道理很硬。多数事情在中间那一段最难看:投入已经不少,回报还没有出现,热情又刚好用完。这时候停下来的人,往往并不是判断出了这条路不对,只是累了。把“半途而废”这四个字放在这里,确实救得了不少人。所以真正的问题不是该不该坚持,而是怎么分清坚持和不肯回头。我的标准有两条。第一条是看方向的证据有没有变。当初决定做这件事,是基于几个判断;如果这些判断今天仍然成立,中间难看只是过程,那就继续。如果其中有一条已经被事实推翻,而你还在往前走,那就不是坚持了。第二条是问一个具体的问题:假如今天才知道这件事,我还会从头开始做吗?回答是会,就继续;回答是不会,那么继续下去的唯一理由就是已经花掉的那些,而它不成为理由。这两条都不好用,因为它们要求人承认自己看错过。可是不用它们,人只能在两句同样古老的话之间来回摇摆:一句让你别停下,一句让你别坚持错的。我因此这样理解这个古老的说法:它提醒的是不要因为累而停,而不是不要因为错而停。前一种停下来值得可惜,后一种停下来其实是及时收手。把这两种混在一起,是这四个字最常被误用的地方。

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

Translation

Part 1 (90–91)

Translate the English passage into Chinese. A Chinese keyboard is required.

Practice — this part is not marked

The real exam asks you to write freely here, and a human examiner reads it. No browser can mark that, and this page does not invent a score: these questions do not count towards your result. Write them anyway — it is the only practice this part allows, and a possible answer is offered once you have handed in.

  1. 90

    The passage to translate

    The mirrors in lifts are not there out of vanity. When tall buildings first appeared, the lifts were slow and passengers complained bitterly about the wait. Engineers found that making the cars faster was expensive, so someone suggested a cheaper answer: give people something to do. Mirrors were put in, and the complaints fell away. The journey took exactly as long as before, but it no longer felt as long. The same trick is used in airports today, where travellers are made to walk further to the baggage hall so that their bags are already waiting when they arrive.

    0 Chinese characters written

    Show a possible answer

    电梯里的镜子并不是为了让人照自己。高楼刚出现的时候,电梯很慢,等电梯的人意见很大。工程人员发现,把电梯做得更快代价很高,于是有人提出了一个便宜得多的办法:给等的人找点事做。镜子装上去以后,意见就少了。路上花的时间和从前完全一样,只是感觉不再那么长。今天的机场用的是同一个办法:让旅客多走一段路才到取行李的地方,这样他们到的时候,行李已经在那里等着了。

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

  2. 91

    The passage to translate

    Hospitals are trying to become quieter. A modern intensive care unit can produce several hundred alarms per patient per day, and studies suggest that the great majority of them require no action at all. The result is what nurses call alarm fatigue: when almost every warning turns out to be nothing, the one that matters is heard as noise. Some hospitals have responded by reducing the number of devices that beep, by sending certain alerts to a screen instead of a loudspeaker, and by letting staff adjust the settings for each patient.

    0 Chinese characters written

    Show a possible answer

    医院正在设法变得安静一些。今天的重病房,平均每个病人每天会响起几百次警报,而研究显示,其中绝大部分根本不需要任何人动手。结果就是护士们所说的那种疲劳:当几乎每一次响声最后都证明没有事,真正要紧的那一次就被当成了背景的声音。一些医院采取的办法是:减少会响的设备的数量,把某些提示送到屏幕上而不是送到喇叭里,并且允许工作人员按每一个病人单独调整。

    This is only one answer among many: yours may be quite different and just as good. It uses nothing beyond your level, which is what proves the task can be done without going outside the word list.

The seven spoken questions, and how to sit them anyway

The real exam has 98 questions. Seven of them are spoken: two of interpreting (92-93) and five of speaking (94-98). This site can neither record nor judge speech, and it will not pretend otherwise: those questions are not in this paper, and your score covers the rest.

The full exam runs 210 minutes, of which about 24 are the 口语 alone; this one runs 180. What follows replaces the difference — not with a written exercise in disguise, but with an examiner you set up yourself.

The speaking paper, ready to copy

Paste this into a fresh conversation with an assistant. It runs the three tasks in the official order, with their real timings, and marks you criterion by criterion at the end. It works typed as well as spoken: typing your answers is already the exercise.

You are the examiner for the speaking paper of HSK7-9 (口语). You run the whole paper on your own. I have no document and no recording: you write them.

Run it in this order, and announce nothing beyond what is asked for.

TASK 1 — 应用读说 (question 94, 3 minutes of speech).
Write, in Chinese, a practical document of about 200 characters: opening hours and rules for a place to visit, a summons, instructions for use, the terms of a registration — your choice, with a table or a numbered list. Then give the instruction in Chinese, on this model: « 你的外籍朋友…… 请你告诉他/她…… (3分钟) ». Then say nothing and wait.

TASK 2 — 听材料回答问题 (questions 95 to 97: 30 seconds, 30 seconds, 2 minutes).
Write, in Chinese, a passage of about 350 characters — a narrative or an exposition — and give it to me as the "material heard". Tell me to read it ONCE ONLY and never go back to it: that is the closest thing to listening when there is no sound. Then ask your three questions in Chinese, ONE AT A TIME, announcing the time allowed for each: two comprehension questions at 30 seconds, then one question asking me to restate or summarise, at 2 minutes. Wait for my answer before asking the next.

TASK 3 — 观点表达 (question 98, 3 minutes).
Ask, in Chinese, an opinion question on a social issue, worded as the official paper words it: « 有人认为……也有人认为……请谈谈你的看法。(3分钟) ».

Rules you hold to throughout:
- Everything you say during the paper is IN CHINESE. Instructions, questions, follow-ups: in Chinese, as on the day.
- You NEVER interrupt me. Until I say or type « 说完了 », you wait. No acknowledgements, no "mhm", no corrections along the way.
- You give me nothing. If I dry up, you wait out the time you announced, then move on.
- You announce the time before each turn. You do not count for me: timing myself is my job.
- If I speak aloud, answer aloud; if I type, type.

Once all five questions are done, and only then, give me the report, IN ENGLISH:
1. A mark out of 5 on each of these five criteria, each with one sentence of justification: faithfulness to the document or the material (did I pass on what mattered without adding anything?), organisation (opening, development, close), range of language (variety of structures and vocabulary, at the level 7-9 expects), accuracy (grammar, measure words, aspect, word order), fluency (pace, hesitation, restarts).
2. For each task, my three costliest errors: the sentence I said, then the sentence a native speaker would say.
3. A full model answer for task 3, at the level HSK7-9 expects.
4. What I should work on first, in one sentence.

Give me no overall mark out of 100 and do not tell me whether I would pass: the HSK 3.0 programme publishes no mark scheme and no pass mark.

Begin now with task 1. Write nothing before the document.

Which assistant

Voice mode is a convenience, not a condition: the prompt works typed anywhere. If you do want to speak, here is what was established on 1 September 2026, and nothing more.

ChatGPT replaced its voice mode with GPT-Live on 8 July 2026: the model listens and speaks at once, so it rides through pauses instead of reading them as the end of your turn — which is what a three-minute answer needs. It does make audible noises of agreement while you speak; the second rule of the prompt asks it to stop. Usable without a subscription, up to a limit OpenAI does not publish.

Claude has a voice mode on mobile, on desktop and in the browser, free accounts included. Its architecture is turn-based: it sometimes pauses on what it takes for the end of a question, which breaks the timing. Anthropic publishes no language list for that mode — open the language picker in your app before counting on it for Mandarin.

On the Chinese side, Qwen (Alibaba) is the only one whose voice mode is reachable from Europe without a Chinese phone number: at chat.qwen.ai, in the browser, with a Google account or an email address — its interface carries a voice-mode button, which we have not tried. Doubao (ByteDance), whose full-duplex model is the fastest measured in 2026 (around 250 ms to resume, announced 9 April), is not distributed on European app stores.

Share this page