What's the Best Anki Deck for Learning Japanese?

Kaishi 1.5k if you are a beginner, a mining setup once you are intermediate. That is the whole answer, and the rest of this page is the detail behind it: what each deck contains, who it is wrong for, and the settings to use once you have imported one.

"What's the best Anki deck for Japanese?" gets asked in every learning community every few weeks, which tells you something. There is no single obvious answer, and the search itself eats time you could have spent learning words. Below is the landscape, plus an alternative worth knowing about: not using a deck at all. (That alternative is our product, so weigh that section accordingly; the deck advice above it stands on its own.)

The best Japanese Anki decks at a glance

Kaishi 1.5k: the best beginner deck

Around 1,500 high-frequency words, one clean card per word, and the current community default for beginners. Ours too.

What is actually on a card: the word with furigana and meaning, an example sentence, and audio for both the word and the sentence. Pitch-accent graphs are included as an optional field, which most beginner decks ignore entirely. The deck is modular, so you can switch fields on and off rather than editing note types by hand, and it has been translated into a dozen languages.

It is actively maintained on GitHub, which matters more than it sounds: most shared decks are a file someone uploaded once and abandoned, and there is no way to tell that from the download page.

Skip it if you already know a few thousand words. It is a starting deck, and starting it late means grinding through hundreds of words you can already read.

The catch: it still lands in your lap as a file with settings you are expected to tune, and it runs dry at 1,500 words with no defined next step.

Core 2k / Core 6k: the old standard

The deck everyone used before Kaishi. Frequency-ordered vocabulary drawn from newspaper corpora, with sentence audio, originally paired with the iKnow course and since re-shared in many forms.

It is still usable, and 6,000 frequency-ordered words is a real corpus. Three problems, though. The card formatting is dated. The versions have forked in confusing ways, so "Core 2k" on one upload is not the same file as "Core 2k" on another, and neither one tells you which you have. And some of the frequency ordering reflects written news Japanese rather than the words you will actually meet first in conversation or in a textbook.

Skip it if you are starting today. Kaishi covers the same ground with better cards and a maintainer.

Tango N5 to N1: sentence-based JLPT decks

The Tango decks are the main alternative shape: instead of a word on the front and a meaning on the back, the card is a full sentence, and you read the whole thing every review.

That shape is the point. You practise words inside the grammar they actually appear in, which is closer to reading than a word-pair card gets. It is also slower per card and harder at the very beginning, because you cannot answer a card until you can parse the sentence around the target word.

Skip it if you are in your first month. Sentence cards assume enough grammar to read the sentence, and if you do not have it yet, every card is four unknowns instead of one.

JLPT vocabulary decks (N5-N1)

Handy if you have an exam date, and the category to be most careful with.

Quality varies enormously between uploads: readings and glosses are often wrong, and the "official" JLPT vocabulary lists these decks claim to follow have not been published since 2010. Everything offered as an official list since then is somebody's reconstruction. That is fine, but it is not the thing the label says it is.

This is the category where a curated corpus is worth the most, because a wrong reading you drill for six months is expensive. If you use one, spot-check twenty cards against a dictionary before you commit a month to it.

Recognition RTK and All in One Kanji

Vocabulary decks teach words; these teach the characters. It is a different job, and you generally want one of each rather than one of the two.

Recognition RTK covers the roughly 2,200 Joyo kanji in the Remembering the Kanji order, focused on recognising a character and recalling its keyword. It is deliberately narrow: no readings, no vocabulary. Fast, and argued about for exactly that reason.

All in One Kanji is the fuller alternative: meaning, on and kun readings, and stroke order on a single card. More information per card, so slower reviews, but you finish knowing how to read the character rather than only how to recognise it.

Skip both if you are already using a kanji-specific app. Doubling up on kanji from two systems is the most common way people end up with 300 reviews a day and quit.

Tofugu hiragana and katakana decks

Two small decks, about a hundred characters each, with audio, stroke order diagrams and mnemonics. There is no real competition here and no decision to agonise over.

Kana is a two-week job, once, at the very start. Use these or any of the free kana courses, get through it, and do not let it become a project.

Mining decks: Yomitan plus a sentence miner

The best long-term setup, and the most work. You read or watch real Japanese, look up words with Yomitan (the maintained successor to Yomichan), and one keypress makes a card from the sentence you actually met the word in.

Why it wins eventually: the cards are words you have already seen, in a context you cared about, which is most of the battle for remembering them. Nothing is arbitrary and nothing runs out.

Why it is not a starting point: you need a few thousand words before mining beats a curated list. Below that, almost every sentence you meet has several unknown words in it, and you end up mining faster than you can review.

Skip it if you are not already reading something in Japanese most days. Mining is a way to keep up with reading, not a substitute for it.

The part the deck recommendations skip

Whichever deck you pick, you inherit the same jobs:

• Vetting it. Shared decks contain wrong readings, missing audio, and glosses that are technically correct and practically misleading. You will not catch these as a beginner, and that's the problem.
• Configuring it. New cards per day, review limits, leech thresholds, FSRS settings, which note fields show on which side. Every one of these has a wrong answer that quietly wastes months.
• Maintaining it. Add-ons break on updates, sync needs setting up, formatting drifts, and the deck eventually runs out.
• Deciding what's next. When the 1.5k is done, you are back where you started, searching forums again.

That's not an argument against Anki. Anki is genuinely excellent and people pass N1 with it. It's an argument about where your first month goes. The most common failure isn't picking the wrong deck. It's spending three evenings comparing decks, importing one, tweaking settings, and never building the daily habit.

How to set up your first deck

If you are importing Kaishi or Core today, these are the settings that matter. Everything else can stay on its default.

• New cards per day: 5 to 10. This is the number people get wrong. Twenty new cards a day feels productive in week one and becomes an hour and a half of reviews by week five, which is when people quit. Ten a day is still thousands of words in a year.
• Maximum reviews per day: leave it high. Capping reviews does not reduce the work, it postpones it into a backlog that gets worse.
• FSRS: on. It is built into modern Anki and it schedules better than the old algorithm with no configuration from you. Turn it on and stop reading about scheduling.
• Leech threshold: 8, action: tag only. A word you have failed eight times needs a different approach, not deletion, and suspending it quietly leaves you thinking you know a word you don't.
• Do the reviews before the new cards. New cards are the optional part of a session. Reviews are the part that decides whether last month counted.

Set these once and do not revisit them. Time spent tuning is time not spent reviewing, and the defaults above are good enough that the difference will never be your bottleneck.

Or: skip the deck entirely

This is the part where we talk about our own product, so read it as the pitch it is. Cotobara is what an already-vetted, already-configured Japanese vocabulary deck looks like when it stops being a file you have to manage:

• Curated, JLPT-organized vocabulary by level and frequency, with no version forks, no dead uploads, no wondering whether the list is current.
• Level-appropriate example sentences on every word, written for the level the word belongs to rather than scraped from a news corpus, with natural-sounding audio, furigana you can switch off, and hover-to-translate glossing.
• A recall-focused review engine that's already tuned. Every word comes back as both a recognition and a production card, and there are no leech thresholds or interval multipliers to guess at.
• Mistake review and visible progress by level and exam-readiness, instead of a card counter.
• WaniKani import, free on every account, so anything you have already burned doesn't come back as a review.

No deck to evaluate, no scheduling settings to research, no maintenance. You sign up and you are reviewing real Japanese in about a minute.

What it costs

The free plan gives you the whole beginner course and reviews for up to 50 words a day, so everything you have learned stays yours to keep studying. Premium is what adds new words - up to 100 a day - plus word breakdowns, notes and bookmarks. Every new account starts with 30 days of Premium, no card required.

Anki is free without a word limit, and that is a genuine point in its favour; what you pay for it is the evening of setup and the ongoing maintenance described above.

Which should you pick?

• Best Anki deck: Kaishi 1.5k for beginners, a mining setup once you're past a few thousand words. Choose this path if you want to build and control your own system: custom card templates, your own algorithm settings, offline data you own forever, free with no word limit.
• If the deck-hunting itself is what's stopping you, try Cotobara: the curation, sentences, audio, and scheduling are already done, and it's free to try, so the comparison costs you a minute rather than an evening.

Common questions

Is Kaishi 1.5k enough for the JLPT N5?

For vocabulary, comfortably. N5 is usually estimated at around 800 words, so 1,500 covers it with room to spare. It will not cover grammar or reading practice, both of which are separately examined, so treat it as one part of an N5 plan rather than the plan.

Should I use Anki or a kanji app for kanji?

Pick one, not both. The two systems will teach overlapping characters on different schedules, and the reviews add up faster than the learning does. If you are already using a kanji-specific app, keep your Anki deck to vocabulary only.

How many new cards a day should I add?

Five to ten. Reviews arrive at roughly ten times your new-card rate once a deck matures, so ten new cards a day settles into a session you can finish, and twenty settles into one most people abandon.

Is Anki free?

On desktop and Android, yes, including sync. The iPhone app is a one-off paid purchase, which funds the rest of the project. Shared decks are free.

Either way, the thing you start this week beats the perfect deck you are still researching next month. For the longer side-by-side, see Cotobara vs Anki, or the wider app landscape if you have not settled on Anki yet. If the setup is the part putting you off, there are alternatives that skip it.

Start for free. No deck hunting, no setup, first review in under a minute.