Every language app eventually gets asked the same question, and most of them dodge it: how many words do I actually need?
The dodge is understandable, because the honest answer has an asterisk on it. But the asterisk is the interesting part, so here is both.
The short answer
Somewhere between 800 and 1,000 high-frequency words will carry you through most ordinary conversation. Not comfortably, not elegantly, but you will be understood and you will understand enough to keep going. That is a much smaller number than the 10,000-word lists that circulate online, and it is small enough to reach in a few months of consistent practice.
The reason is that word frequency in every language is brutally lopsided. A small set of words does an enormous share of the work - the articles, the pronouns, the twenty verbs you use constantly, the connectives that hold sentences together. Learn those and you can say things. Learn 4,000 rarer nouns instead and you can name a lot of objects while being unable to ask where one is.
The asterisk: which words, in which order
This is where most vocabulary advice goes wrong. "How many words" is the wrong question in the same way "how many miles" is the wrong question about a route. Two learners with 900 words each can be in completely different places depending on which 900.
So the number that matters is not the size of the list. It is:
- Frequency order. The first hundred words should be the hundred you cannot
avoid using. Every level of our decks is built frequency-first, so the early cards are load-bearing rather than decorative.
- Production, not recognition. Recognising bibliothèque on a page is a
fraction of the work of retrieving it when you need it. Typed recall is the default on Semantix for exactly this reason: reading an answer and deciding you knew it is the easiest way to fool yourself.
- Words attached to a sentence. A word learned alone is a word you can
translate. A word learned inside a sentence you understood is a word you can use. Every vocabulary card we ship carries an example sentence, and most carry a fill-in-the-blank twin built from it.
The actual numbers we ship
We would rather give you the real figures than a marketing round number. Across 19 languages, the authored vocabulary is 7,335 cards, banded by CEFR level:
| Level | Cards per language | What it is for | |---|---|---| | A1 | 106-157 | Survival: greetings, numbers, family, food, the present tense | | A2 | 74-80 | Everyday life: past and future, directions, work, plans | | B1 | 60-63 | Opinions, comparisons, abstract-but-common vocabulary | | B2 | 30-67 | Nuance, register, the vocabulary of argument | | C1 | 60-63 | Precision: near-synonyms, idiom, low-frequency but exact words |
Two things about that table are worth calling out, because they look like flaws until you know why.
The levels get smaller, not bigger. That is deliberate. A1 has to build a foundation from nothing, so it is the biggest band by a wide margin. By C1 you are not learning categories of thing any more, you are learning the difference between two words that both mean "to fix" - and that work is far more effective in small, carefully chosen batches than in bulk.
Those numbers are the authored floor, not your ceiling. The decks are the scaffold. The vocabulary that actually sticks tends to be the words you met in your own conversation, wanted, and saved - and those cards go into the same spaced-repetition schedule as everything else. A learner six months in usually has more of their own words than ours.
Why we cap what you get per day
Semantix releases new vocabulary in a small daily batch rather than dumping a level on you. And when your review queue is already backed up, the drip pauses rather than adding to it.
This is the single most-requested thing to turn off, and the single thing we are most confident about keeping. A vocabulary app's failure mode is not "too little" - it is a queue of four hundred cards that grows faster than you can clear it, at which point a learner does the rational thing and stops opening the app. The number of words you have seen is not a measure of anything. The number you can retrieve is.
So: how many?
Aim for 1,000 words you can produce before you worry about the second thousand. At the rate Semantix releases them, that is a few months of ordinary practice - and by the time you get there you will have had hundreds of real conversations in the language, which is where the words go from known to yours.
If you want to see the actual list for a language before you start, every one of the 19 has a page: what it covers, how it is levelled, and what the first unit teaches.