How long does it take to learn Cantonese?
The US Foreign Service Institute reaches professional working proficiency in Cantonese in about 2,200 class hours — its hardest band, alongside Mandarin, Japanese, Korean and Arabic, roughly four times Spanish. At an hour a day with a tutor, conversational Cantonese (B1) takes around four and a half years. Six tones and a character set shared with Mandarin are the obvious costs. The underrated one is that spoken Cantonese has no fully standardised written form, so the usual strategy of learning through reading works less well than in any other language here.
- FSI baseline
- 2200 class hours
- Difficulty band
- super hard
- FSI category
- Category IV / Category V
Cantonese sits in the Foreign Service Institute's hardest band: 2,200 class hours to Speaking-3 / Reading-3 — alongside Mandarin, Japanese, Korean and Arabic, and roughly four times what Spanish takes.
FSI lists it as its own language with its own figure, not as a variant of Mandarin.
The two standing warnings:
Those are classroom hours, not app minutes. Full-time small-group instruction with qualified teachers.
2,200 hours is professional proficiency — arguing a point in your field, reading a newspaper unaided.
Six tones, and why they are harder than four
Mandarin has four tones distinguished largely by contour — a rise, a fall, a dip. Cantonese has six, and several of them are level tones at different pitch heights.
That is a subtler perceptual task. Telling a rise from a fall is a shape comparison. Telling three flat tones apart means judging absolute pitch relative to a speaker's range, and that range differs between speakers.
Mandarin speakers learning Cantonese consistently name the tone system as the main obstacle, despite already treating pitch as lexical. If someone with tonal experience finds it hard, budget accordingly.
The problem that is specific to Cantonese
Here is the structural fact that makes Cantonese different from every other language on this site, including Mandarin.
Written standard Chinese is based on Mandarin. Its grammar and vocabulary follow Mandarin, not Cantonese. A literate Cantonese speaker reads that standard, pronouncing the characters in Cantonese — but what is on the page is not a transcription of how they speak.
There is a written Cantonese, used in informal contexts, with characters specific to it. It is not fully standardised, and it is not what formal materials teach.
The consequence for a learner is that the usual strategy of learning a language through reading works less well here than anywhere else on this site. In Spanish or Russian, reading and speaking reinforce each other directly. In Cantonese, reading builds a formal register that does not map onto conversation.
You have to acquire the spoken language largely through listening and speaking, which is slower and requires people.
Where the time actually goes
Characters, the same open-ended cost as Mandarin
Thousands of items, each with a form, a meaning and a Cantonese reading, acquired continuously rather than finished. Same as Mandarin in size, with different pronunciations attached.
If you already read Chinese, this transfers substantially and is the largest single thing you bring.
Material scarcity, which is not difficulty but behaves like it
Far fewer graded readers, apps, structured courses and tutors than Mandarin, because publishing follows learner numbers.
Our model prices study time and cannot see this. The practical effect is that the cheap self-study route is weaker in Cantonese than in almost any language here, and you will need to pay for a person earlier in the project than the hour figures suggest. That is a real cost the 2,200 does not contain.
What each level actually costs
| Goal | Method | Estimate |
|---|---|---|
| A2 — simple everyday exchanges | App, 20 min/day | around 1,099 hours, ~9 years |
| B1 — hold a conversation | Tutor, 1 hr/day, 5 days | around 1,155 hours, ~4.4 years |
| B2 — work, study, argue a point | Tutor, 1 hr/day, 5 days | around 1,686 hours, ~6.5 years |
| C1 — professional proficiency | Intensive class, FSI-style | 2,200 hours, ~20 months full-time |
If you already speak Mandarin
The characters and a large share of formal vocabulary transfer, and that is a genuine head start on the reading half.
The spoken languages are not mutually intelligible. The tone systems are different, the everyday vocabulary differs, and the grammar differs in places. A Mandarin speaker starting Cantonese has done the writing system and not the language.
Our model treats them as separate languages in the same branch, with a real but partial discount — which is the honest shape of that relationship.
What actually works
- Learn jyutping early. It is standardised, it encodes tone explicitly, and it makes dictionaries usable — unlike the romanisations for several other languages here.
- Do not rely on reading to build speech. This is the language where that strategy transfers least, and knowing it in advance changes how you allocate hours.
- Budget for a tutor sooner. Material scarcity is the practical difficulty, and a person substitutes for the graded readers that do not exist.
- Drill the level tones against each other, not in isolation. The difficulty is relative pitch, and isolated practice does not build it.
How you study matters more than which language
The same target, the same hour a day, five different ways of spending it.
Common questions
- How many tones does Cantonese have compared with Mandarin?
- Six, against Mandarin\u2019s four, and they are distributed differently — Cantonese distinguishes several level tones at different pitches where Mandarin relies more on contour. The practical effect is that Cantonese tones are harder to tell apart for a beginner, because distinguishing three flat tones by pitch height is a subtler perceptual task than distinguishing a rise from a fall. Mandarin speakers learning Cantonese report the tone system as the main obstacle, despite already being used to lexical tone.
- Can I read Chinese if I learn Cantonese?
- You can read written standard Chinese, because that is what is taught and what is printed — but written standard Chinese is based on Mandarin grammar and vocabulary, not on how Cantonese is spoken. So a literate Cantonese speaker reads a newspaper that does not represent their speech, and reads characters aloud with Cantonese pronunciations. There is a written Cantonese used in informal contexts with characters specific to it, and it is not fully standardised and not what formal materials teach.
- Why are there so few Cantonese learning materials?
- Because course publishing follows learner demand, and Mandarin has vastly more learners for reasons of policy and scale. The practical consequence for you is not difficulty but method: fewer graded readers, fewer apps, fewer structured courses, and more reliance on tutors and native material. Our model prices study time and cannot see material scarcity, so the real-world project is harder than 2,200 hours implies in a way that shows up as needing to pay for a person sooner.
- Should I learn Mandarin first and then Cantonese?
- Only if you want Mandarin for its own sake. They are separate languages priced separately at 2,200 hours each, and while characters and a large share of formal vocabulary transfer, the spoken languages are not mutually intelligible and the tone systems are different. Learning Mandarin first buys you the writing system and a formal vocabulary layer, which is real, and it does not shorten the spoken half of Cantonese. If your reason is Hong Kong or a Cantonese-speaking family, start with Cantonese.
- What is jyutping and do I need it?
- A romanisation system for Cantonese that marks tones with numbers, widely used in dictionaries and teaching materials. Yes, learn it — unlike Thai romanisation it is standardised and it encodes tone explicitly, so it does the job a phonetic notation should. Use it as a pronunciation aid and a dictionary key alongside characters, not as a replacement for them, because everything written for adults uses characters and jyutping never appears outside learning contexts.
If you already speak one of these
Our estimate applies a reduction for each, and each has its own write-up.
- Mandarin Chinese — cuts roughly 30% off
- Japanese — cuts roughly 15% off
Where to study Cantonese
Matched by the same rules the calculator uses, at a B1 target with no prior related language. The rules were written before any of these were added and the ranking code cannot read commission rates — a test rewrites every payout and checks the order does not move.
Anyone who needs to retain vocabulary — free, and better at spaced repetition than every paid app.
Not for: People who want a curriculum — Anki is a tool, not a course
Intermediate learners who want native input from video with dual subtitles, at no cost.
Not for: Beginners — unscripted native speech below A2 is noise, not input
- Rocket Languagesabout $260 once · Structured self-study course · up to B2
Learners who want one structured course covering grammar explicitly, bought once rather than rented.
Not for: Anyone who wants live correction from a human — this is a recorded course
- Pimsleurabout $150 once · Structured self-study course · up to B1
Learners who want to speak from the first lesson and can give it thirty focused minutes of audio a day, hands-free.
Not for: Anyone who wants to read or write the language — the method is almost entirely spoken
Your own target and budget change this list — run it for your case. Prices carry the date they were checked and go stale; confirm on the merchant's own site before paying.
Sources
Where the baseline figure on this page comes from. Everything applied on top of it is our own estimate.
- Foreign Language Training — US Department of State, Foreign Service Institute
The class-hour bands every estimate on this site starts from, and the definition of the proficiency they lead to (ILR Speaking-3 / Reading-3).
- FSI Language Difficulty Ranking — fsi-language-courses.org
Which languages sit in which band. The State Department publishes the categories as an image, so per-language assignment comes from independent reproductions that agree with each other and with the official hour bands.
By Kevin Fan. Native Mandarin speaker who learned English as a second language — the same pairing this site prices at 2,200 hours, done from the far side. Builds and maintains the estimate model and every page on it. Last reviewed 2026-08-08. How we work these numbers out, including which coefficients are our own estimates.