The 5 best Korean Anki decks, with the numbers nobody checks

Here is the moment this page is written for. You have decided to learn Korean with Anki, you search for a deck, and every list hands you the same three names with no numbers attached. So you download the biggest one, study for two weeks, and discover that most of the cards have no audio.

That is avoidable, and the fix takes thirty seconds if you know where to look. AnkiWeb prints a note count and an audio count on every deck page, and the gap between them tells you what you are actually getting. The figures below were read off those pages in August 2026, not copied from another blog post.

One disclosure before the list. We make a Korean course, so treat the final section accordingly. Everything before it is a straight comparison of what these decks contain.

AnkiWeb reports notes and audio files as separate counts, and the gap between them is the single most useful thing on a deck page. A deck can list 6,500 notes and 831 audio files, which means seven cards in eight are silent. Figures below were read from each deck's AnkiWeb page in August 2026.

Korean Vocabulary by Evita is the deck people mean when they say “the Anki deck”. It holds 6,500 words and it is the most rated Korean deck on AnkiWeb. The ordering is worth understanding: it follows how many stars each word carries in the National Institute of Korean Language's dictionary, so it is a usefulness ranking rather than a raw corpus count. Words of Chinese origin carry their hanja, which you can ignore or use depending on whether you have studied Chinese or Japanese.

The catch is audio. The deck lists 831 audio files against 6,500 notes, so most cards are silent, and its last update was May 2022. The author has also noted that the deck may not stay free indefinitely. None of that makes it a bad deck. It makes it an incomplete one, and reviewers on the deck page say the same.

Korean Core Vocabulary [Audio + Sentences] is the interesting one, because it fixes exactly that. It is explicitly based on Evita's deck, carries the same 6,500 words, and adds an audio file for every single note plus an example sentence and hanja. It was updated in December 2025. The tradeoff is that it is new enough to have almost no ratings, so you are trusting the audio quality rather than a decade of feedback.

If you want a much bigger net, Korean Vocabulary Core 2k/5k/10k by Soi runs to over 45,000 notes across several languages. It has no audio at all, and 45,000 cards is not a beginner decision. Mentioned for completeness rather than as a recommendation.

Korean Grammar Sentences by Evita is the one deck here with no real competitor. Roughly 2,450 notes, 2,335 of them with audio, sorted into seven difficulty levels you can suspend individually. The sentences teach grammar patterns in context rather than as isolated rules, and the deck description maps each level to the constructions it introduces.

Two honest notes. Levels six and seven were still unfinished at the last update in February 2021, and the author states the deck covers enough grammar for TOPIK level 4, not beyond. For a beginner working through sentence patterns, that is more than enough. Our sentence structure guide covers the word order those sentences are built on.

TOPIK 1 Korean Vocabulary is the most recently maintained deck in this list, updated in January 2026. It holds 1,103 words with an audio file for each, plus part of speech, romanization and example sentences. If you have a TOPIK I date booked, a focused eleven hundred words beats a 6,500 word deck you will not finish first. The guide to learning Korean vocabulary covers how to pace that against a deadline.

TTMIK's First 500 Korean Words by Retro pairs 500 words with 500 audio files and 361 pictures, which makes it a gentle first deck. Read the description before committing: the author has moved the current versions to a Google Drive folder and left the AnkiWeb copy in place as a signpost, so what you download here is a 2021 snapshot.

Anki ships with defaults tuned for nobody in particular, and three changes do most of the work.

Turn on FSRS. It schedules from your actual review history rather than the older fixed multipliers, and it is built into current Anki versions. Set desired retention to about 0.9. Higher sounds better and simply buys you more reviews per day for a marginal gain in recall.

Cap new cards at 10 to 20 a day. This one setting decides whether you are still using the deck in March, and the failure runs on a predictable timetable. Sunday you feel keen and add 100 new cards. Monday is fine. Wednesday there are 180 reviews waiting and you do half. Thursday there are 240, because the half you skipped came back with the new ones. By the weekend the number is unrecoverable and you stop opening the app. That is the story behind almost every “Anki did not work for me” post, and the cause is not the software.

Add a typing field. Most Korean decks ship recognition only, showing you the word and asking whether you remember it. Anki supports typed answers with {{type:Korean}} on the card template. This matters more in Korean than in most languages, because spelling preserves letters that pronunciation drops: you can recognise 좋아요 for months and still write 조아요. The pronunciation rules guide covers why that gap opens.

A shared deck hands you somebody else's list and stops there. Three things it will not do, in rough order of how much they cost you.

It will not tell you which card you got wrong and why. Anki marks the whole answer right or wrong, so a missing ㅎ and a completely different word score the same. It will not keep itself current, which is why the update dates above matter. And it will not decide anything for you: which deck, which settings, whether to suspend a level, whether today's 200 reviews are worth doing.

For some people that control is the entire appeal, and those people should use Anki and enjoy it. If you have abandoned a deck before because the evening went on configuration rather than Korean, the useful conclusion is about you rather than about the software. Our look at what exists for Korean covers that split in more detail.

Back to the person at the top of this page, two weeks into a silent deck. The thirty seconds that would have saved them was reading two numbers on the deck page before downloading. Notes, then audio. If the second is much smaller than the first, you are choosing a reading deck, which is a fine thing to choose on purpose and a miserable thing to discover in week three.

This is a real review card from the course, not a screenshot. Type it with the on-screen keyboard or your own, and get one wrong on purpose to see the letter-level feedback a shared deck cannot give you.