Ask which characters to learn first and you will be given a list. Ask where the list came from and you find there are at least four of them, built for different purposes, and they disagree about the first hundred characters more than most learners would expect. Picking the wrong one for your goal costs months.
Four orderings, four different jobs
Frequency order comes from counting characters in a body of text and sorting by how often they appear. It gives you the most reading coverage for the fewest characters learned — ideal if your main goal is to read. The trade-off is that you will learn characters you cannot write, pronounce in isolation, or use in a sentence, since frequency has nothing to say about usability.
The national standard list — the 通用规范汉字表 of 2013 — sorts 8,105 characters into three tiers. Its purpose is specific: it governs publishing and information processing by defining which characters a system or publisher must handle. It was never designed as a syllabus, and using it as one is borrowing a tool from another trade.
The exam ordering follows the HSK proficiency syllabus, which sets out what a candidate is expected to know at each band. It is built backwards from an assessment, which makes it excellent if you are sitting that assessment and arbitrary if you are not.
Component order teaches characters in the order that lets each new one be built from parts you already know. It is the strongest for retention and the weakest for early reading, because the characters that decompose most cleanly are not the characters that appear most often.
The split the exam authority does make
The current HSK standard makes one useful distinction that is not often reported.
The syllabus separates characters you are expected to recognise from characters you are expected to handwrite, and the two sets are different sizes: about 3,000 recognised against 1,200 written at the top of the scale. That is an examining body stating, in its own materials, that reading order and writing order are not the same problem — which is precisely the distinction the four orderings above blur.
What it does not say, despite being widely repeated, is that handwriting is deferred to the upper bands. The standard sets a handwriting requirement from the first band — a hundred characters at band one, two hundred at band two, three hundred at band three — with its own separate list. If you have been told you need not write anything until level five, you have been misinformed; the standard says otherwise.
Two things about HSK 3.0 that are commonly stated wrongly
The standard behind the current syllabus has been in force since 2021. The examination is what changed more recently, and the two get merged constantly, usually in the direction of describing a five-year-old standard as new.
The rollout is also less finished than most coverage suggests. As of writing, the new exam is still described as being in trial: a first international trial sitting earlier in 2026, another scheduled for September, and no announced date for full replacement. Claims of a specific completed switchover date circulate on commercial preparation sites and are not on the examining body's own notice board.
How much the orderings actually disagree
To compare them fairly, take the first hundred characters from each list and measure the overlap. Our characters by learning order tool follows one of these orderings, and the stroke order animator covers the writing half. Do that and the lists share far less than the confident tone of any single list would suggest. A character in the first hundred by frequency may be several hundred places down a component-ordered list, because it is common but structurally complex. A character early in a component ordering may be rare enough that you will not meet it in text for months.
This is not a defect in any of them. Each is optimal for its own objective and none is optimising for the others. The mistake is treating a list built for one purpose as though it were the canonical order.
Choosing on purpose
If the goal is reading — signage, menus, subtitles, a news headline — frequency order gets you there fastest, and you should expect to be unable to write most of what you can read. It is a reasonable trade-off, and better to make it deliberately than to discover it by accident.
If the goal is an exam, use that exam's own list, because anything else optimises for a target you are not being measured against.
If the goal is durable memory, component order costs more early and pays back later, since each new character lands in a structure rather than as a fresh shape.
And if the goal is handwriting specifically, note that the recognise-versus-write split means the writing list is the smaller one. Learning to write everything you can read is a much larger project than the syllabus asks for.
Where this comes from, and what will date
The tier counts are from the 2013 national character standard, and the band structure and the recognise-versus-handwrite split from the current proficiency standard in force since 2021. Per-band figures should be taken from the standard itself rather than from preparation sites, which is where the two claims corrected above originate.
The examination timetable is the part that dates, and it is currently in motion, so the position described here is as at the time of writing rather than settled. The four orderings and the reasons they differ are structural and will outlast any particular syllabus revision.
