How Many German Words for B1, B2 and C1?

The honest answer: roughly 650 words for A1, 1,300 for A2, 2,400 for B1, about 4,500 for B2 and 8,000-9,000 for C1. The first three of those are not estimates - they are the sizes of the official Goethe-Institut and ÖSD exam word lists, which you can download for free. The last two are estimates, because no exam board publishes a list above B1 and the CEFR itself specifies no vocabulary size at any level. Anyone who tells you "B2 is exactly 5,000 words" is quoting a guess that someone else made confidently.

What makes the guesses useful is frequency coverage: how much of an ordinary German page you can read if you know the top N most common words. Below we publish our own numbers for that, worked out from the frequency bands behind the merke.me vocabulary decks, and they change the picture. Going from 0 to 2,000 words buys you about 80% of a normal page. Going from 2,000 to 5,000 buys you another 8 points. The curve flattens hard, and knowing where it flattens is more useful for planning your study than any single target number.

The numbers by level

Two things to notice. First, the reliable numbers stop at B1. The Goethe-Institut deliberately stopped publishing word lists above that level because at B2 and beyond the exam tests what you can do with unfamiliar text, not whether you memorised a closed set. Second, the jump from B1 to C1 is roughly three times the entire A1-B1 vocabulary. Most learners underestimate this by a wide margin and then wonder why B2 takes eighteen months.

Why the CEFR itself gives you no number

The Common European Framework of Reference is a description of what you can do - order a meal, follow a lecture, argue a position in writing. It contains can-do descriptors, not word counts. Search the full document for a vocabulary target and you will find only scales like "has a good range of vocabulary for matters connected to his/her field" for B2. That is deliberate: a Turkish learner and a Dutch learner reach the same can-do statements with very different word inventories.

So every number in circulation traces back to one of three sources:

• Exam word lists. Authoritative but only up to B1, and specific to one exam board. The Goethe A1, A2 and B1 lists are free PDFs and are the ground truth if you are sitting that exam.
• Textbook corpora. Researchers count the distinct lemmas introduced across a coursebook series up to a given level. This is where most B2 figures come from, and different series disagree by 30% or more.
• Frequency coverage. You pick a comprehension threshold - typically 95% or 98% of running text - and ask how many words you need to hit it. This is the most transferable method and the one we can contribute data to.

Original data: what each frequency band covers

The table below comes from our own frequency ranking, built from a mixed contemporary corpus of about 220 million running words: news, film and TV subtitles, published fiction, Wikipedia and forum threads. Text is lemmatised, so ging, gehe and gegangen all count once as gehen. Proper nouns, numerals and pure interjections are excluded. "Unknown tokens per page" assumes a 350-word page and counts repeats, which is what your reading experience actually feels like.

The research consensus is that comfortable independent reading starts around 98% coverage and that 95% is the floor for guessing unknown words from context. On these numbers, German does not reach 95% coverage until somewhere past 16,000 lemmas, and 98% is out beyond 40,000. That sounds bleak until you notice the other half of the picture: the first 2,000 words do four-fifths of the work, and they cost maybe 250 hours of honest study. The next 3,000 words buy 8 percentage points. The 5,000 after that buy 4.5.

Coverage also depends heavily on what you read. Same bands, split by text type:

This is why learners who watch a lot of German TV feel fluent and then open the Süddeutsche and feel like beginners. At 2,000 words you are 6 points better off in dialogue than in news, which is the difference between roughly 50 and roughly 82 unknown tokens per page.

Why German counts look worse than English ones

If you have seen English coverage figures, German's look disappointing at the same rank. Three reasons, and they are not all bad news.

• Compounding. Krankenversicherungsbeitrag is one token and one lemma, and it will never make a top-10,000 list. But if you know krank, Versicherung and Beitrag, you read it fine. Raw coverage numbers systematically understate what a German learner actually understands.
• Inflection and separable verbs. Lemmatisers disagree about whether anrufen used as rufe ... an is one lemma. Different published counts use different rules, which is a large part of why estimates for B2 range from 3,500 to 6,000.
• Register split. German maintains a sharper divide between spoken and formal written vocabulary than English does, so the same word count gets you further in conversation and less far in print.

The practical consequence: chase word families and building blocks rather than raw entries, and spend real time on how words combine. Knowing Entscheidung is a third of the job; knowing eine Entscheidung treffen is the rest, which is why we keep collocations as their own layer rather than folding them into single-word cards.

Where word counts mislead you

A count is a planning tool, not a goal. Three places it goes wrong.

Passive versus active. The C1 figure of 8,000-9,000 is recognition vocabulary. Your productive range at C1 is realistically 3,000-4,000 words you can retrieve under time pressure. Speaking exams test the smaller number. If you are failing speaking sections while your reading is fine, adding vocabulary is the wrong fix - drilling retrieval is.

Grammar is not vocabulary. No word count gets you through B2 Schreiben if your Konjunktiv II and Passiv are unreliable. That is a separate track, and a smaller one - the grammar library is finite in a way that vocabulary is not.

If you have an exam in six weeks, use the exam list. The Goethe-Institut publishes the A1, A2 and B1 Wortlisten as free PDFs. They are authoritative, they are exactly what the item writers used, and no third-party deck - ours included - is a better match for that specific test. Download the PDF. If you like building your own cards and don't mind an evening of setup, Anki plus that list costs nothing and works. Our case is for the learner who wants the sequencing, audio and example sentences done for them, and who is going past B1 into territory where no official list exists.

Start with the 2,000 that pay first

If you take one thing from the table: the top 2,000 lemmas are worth 79.6% of an average page and cost a fraction of what the next 3,000 cost. Learn them in frequency order, with audio and in sentences, and you cross from decoding to reading faster than any other route.

That is exactly how our vocabulary decks are ordered - frequency bands first, exam lists mapped on top, spaced repetition handling the review schedule so you are not re-reading words you already know. Browse the word entries to see the format before committing, or check pricing if you have already decided.