JLPT N5 Vocabulary: The List That Isn't Official

Search for an N5 vocabulary list and the pages disagree about the most basic thing: how many words there are. Most say around 800. Some say 700-something, a few push past 1,000, and none of them flag the disagreement. Usual disclosure: I build Piccard, a flashcard tool whose deepest language support is Japanese, so the list you print matters less to me than what happens to those words once the novelty of printing wears off.

Why no two N5 lists agree

The reason is not sloppy editing. Since the 2010 revision, the JLPT has published no test-content specification at all; the organizers' stated position is that studying from kanji and vocabulary lists is discouraged, so there is nothing official to copy from (jlpt.jp; Wikipedia). Every list you can download is an inference.

There used to be a real one. Until 2009, the old-format test published full kanji, vocabulary, and grammar lists per level; old Level 4, N5's ancestor, ran to about 100 kanji and roughly 800 words (728 is the most-cited reconstruction). Even then, by most accounts roughly 20 percent of any exam's vocabulary could sit outside the published list, and the list described a format that stopped being administered that year. Today's N5 lists each start from that old base and add their own guess at fifteen years of drift, which is why they cluster near 800 and never match.

One useful property survives the disagreement: the core is stable. Fourteen words that sit on essentially every reconstruction:

Plenty of N5 vocabulary (でも, とても, ちょっと) is kana-only and fills its own reading column; the rows that need the column are exactly the ones carrying kanji.

What the exam actually asks

On N5 the paper has three sections, and vocabulary is the shortest: 25 minutes against 50 for grammar-and-reading and 30 for listening, 105 minutes in all. The total score is 180, and passing takes 80 of it (44.44 percent), with a floor on each section, 38 points on the combined language-knowledge section and 19 on listening. The exam is given twice a year in Japan and selected countries, on the first Sunday of July and December, and once a year elsewhere; check your local administrator's site for which applies to you.

Vocabulary is a 25-minute section, but it is load-bearing in all 105: the words appear in the vocab questions, again inside every reading passage, and a third time as audio, where the kanji are gone and only the reading you attached to the word survives. A list row that pairs 電車 with "train" but never with でんしゃ is preparing you for one third of the paper. Recognition of a printed pair is not retrieval of a spoken word; the gap between the two is the testing effect doing its quiet damage.

The list you download, the list you build

The downloaded list is a starting asset, not a plan. Here is how I would feed it through Piccard, Japanese direction specifically.

Rank before you study. Upload the list as a pdf, or a phone photo of the printed page, and every word comes back as a candidate card scored 0–100 for how much it is worth learning. Words you already half-know from anime or a semester of class sink to the bottom and get skipped in one pass. A single confirmation turns up to 20 of them into cards. The ceiling lives in our code as a named constant, set because the model spends roughly 300–350 output tokens per card. Japanese cards carry the reading on the card, so 食べる arrives as 食べる（たべる）, not naked kanji.

Then let material take over. On a captioned YouTube video in Japanese, click one subtitle line and candidates come back from that line only. A video without captions has nothing to click, and the extension says so instead of improvising. A double-click on any Japanese word on any web page opens a popup dictionary backed by JMdict. Cards saved from a video keep their source: when the card comes due, the line, the voice, and the scene replay. File-based transcription (dropping an mp3 or mp4 in for processing) is built for English audio only; Japanese listening material reaches your deck through the caption route instead.

Then the schedule does the work

Eight hundred rows defeat themselves by week three: the first hundred words fade while you memorize the second hundred. The fade is no mystery; it was measured a century ago and named the forgetting curve, and the answer to it, spaced repetition, is older than every app that ships it. Review each item just before it drops, with the gap growing every time you survive, and each cycle makes the memory cheaper to keep. Recall itself, not re-reading, strengthens the trace; a paper list you hover your eyes over is barely studying at all.

Scheduling in Piccard is FSRS-6 with zero configuration: 0.9 target retention, four ratings (again, hard, good, easy), and no settings page where a beginner could quietly set it wrong. The plan at N5 scale:

1. Start at 10 new cards a day from the ranked list. Empty the review queue every day, no exceptions.
2. Around week two, captioned videos should pass the list as your card source.
3. When a card comes up, say the reading aloud before you flip. The listening section grades exactly this move.
4. Count backward from your sitting. At 10 new cards a day, 800 words is about 80 days of intake; the first-Sunday date is fixed, so the final weeks should be pure review.

If your next question is "then which deck?", the frequency-first answer has its own write-up: Kaishi 1.5k: What the Default Japanese Starter Deck Is, N5-and-beyond vocabulary ordered by usage, not by exam. The Best Anki Decks for Japanese compares the rest of the deck landscape. And the sister problem, an exam that publishes two competing lists instead of none, is HSK 1 Vocabulary: 150 Words or 500?.

The short version

No official N5 list exists under the current format. The old 728-word spec, the Level 4 list it was reconstructed from, died with the 2009 format change; everything since is inference. Pick any reputable list, treat it as raw material, and put your trust in the schedule instead: rank out what you already know, keep readings attached, run into the rest in real sentences, and let review dates run backward from the sitting you registered for.

The free tier is 150 energy (about 150 cards, roughly a fifth of an N5 list, or the whole thing minus what you already know), then prepaid top-ups at $5 / $10 / $20; no subscription, and topped-up energy never expires. To start: Piccard on the Chrome Web Store.

All guides: Piccard Blog.