The Common European Framework of Reference for Languages sorts language ability into six levels, from A1 to C2, using can-do statements rather than grammar checklists. This guide reads those levels through one skill: speaking. The official descriptors for spoken interaction and spoken production, the vocabulary each level implies, and why your own guess about your speaking level is the least reliable number in the whole system.

Most CEFR guides give you a tidy four-skill summary and then send you to a placement test. This one inverts the question. Instead of asking what an A2 or a B2 knows, it asks what happens when a person at each level opens their mouth in a real conversation, and what a learner should stop doing if they want to move up.

What this guide covers

What the CEFR actually is, and what it is not

The CEFR is a scale of behavior, published by the Council of Europe, that describes what a person can do with a language. It has six levels grouped into three bands: A1 and A2 form the Basic User band, B1 and B2 the Independent User band, C1 and C2 the Proficient User band. The 2020 Companion Volume added a Pre-A1 stage below A1 for learners who have a repertoire of memorized words and phrases but cannot yet build their own sentences.

Start with what the CEFR is not, because that is where most confusion comes from.

It is not a course. No level tells you which tenses to study or in what order. It is not a test, either. Cambridge, IELTS, DELE, TestDaF, and the rest map their scores onto CEFR levels, but the framework itself contains no exam. It is not a measure of vocabulary size, and it is not a number of hours. And it is deliberately not a single score. The official CEFR self-assessment grid published by Europass and the Council of Europe describes five separate abilities at every level: listening, reading, spoken interaction, spoken production, and writing.

That last point is the one worth holding onto. The framework expects you to sit at different levels in different skills. Reading at B2 while speaking at A2 is not a contradiction in CEFR terms. It is the single most common shape a learner has, and the framework was built to describe exactly that. We wrote about the mechanism behind it in why you can understand a language but can’t speak it.

So a sentence like “I am a B1” is almost always too coarse to be useful. The useful sentence names the skill: “my spoken interaction is around B1.”

The six levels by what you can say

Below are the official spoken interaction and spoken production descriptors for all six levels, quoted verbatim from the Council of Europe and Europass self-assessment grid, followed by a plain reading of what each one means in a real conversation.

Note the split. Spoken interaction is talking with somebody: turn-taking, answering, repairing, keeping up. Spoken production is talking at somebody: describing, explaining, presenting, telling a story without interruption. They move at different speeds. Most learners are better at production than interaction, because production lets you plan and interaction does not.

A1: someone has to meet you halfway

Spoken interaction. “I can interact in a simple way provided the other person is prepared to repeat or rephrase things at a slower rate of speech and help me formulate what I’m trying to say. I can ask and answer simple questions in areas of immediate need or on very familiar topics.”

Spoken production. “I can use simple phrases and sentences to describe where I live and people I know.”

The defining feature of A1 is the conditional clause: provided the other person is prepared to. At A1 the conversation only works if your partner cooperates actively, slowing down, repeating, and finishing your sentences for you. You can ask where the station is and say where you live. You cannot yet carry your half of the exchange without help.

A2: routine exchanges, but you cannot keep them going

Spoken interaction. “I can communicate in simple and routine tasks requiring a simple and direct exchange of information on familiar topics and activities. I can handle very short social exchanges, even though I can’t usually understand enough to keep the conversation going myself.”

Spoken production. “I can use a series of phrases and sentences to describe in simple terms my family and other people, living conditions, my educational background and my present or most recent job.”

A2 is where transactions work. Ordering, buying, asking directions, checking in, exchanging basic personal information. The honest limit is written straight into the descriptor: you cannot usually keep the conversation going yourself. Small talk starts and then stalls, because the moment the topic drifts off the script you run out of material. This is the level most people plateau at after a school course, and it is the one that feels most frustrating, because you can clearly do things with the language and still cannot have a conversation.

B1: the first level that is genuinely conversational

Spoken interaction. “I can deal with most situations likely to arise whilst travelling in an area where the language is spoken. I can enter unprepared into conversation on topics that are familiar, of personal interest or pertinent to everyday life (e.g. family, hobbies, work, travel and current events).”

Spoken production. “I can connect phrases in a simple way in order to describe experiences and events, my dreams, hopes and ambitions. I can briefly give reasons and explanations for opinions and plans. I can narrate a story or relate the plot of a book or film and describe my reactions.”

Two words carry B1: enter unprepared. This is the first level where you can start a conversation you did not rehearse and stay in it. You can tell a story, give a reason, describe a plan. You will be slow, you will circle around missing words, and you will make errors constantly, but the conversation survives. If your goal is to travel, make friends, and be a functioning adult in the language, B1 is the door.

B2: your partner stops working for you

Spoken interaction. “I can interact with a degree of fluency and spontaneity that makes regular interaction with native speakers quite possible. I can take an active part in discussion in familiar contexts, accounting for and sustaining my views.”

Spoken production. “I can present clear, detailed descriptions on a wide range of subjects related to my field of interest. I can explain a viewpoint on a topical issue giving the advantages and disadvantages of various options.”

B2 is the level where the burden shifts off the other person. Below B2, native speakers adjust for you, consciously or not. At B2 they stop having to. You can argue a position and hold it under pushback, which is a genuinely different skill from describing your weekend. This is the level most employers and universities mean when they say “fluent,” and it is where the FSI’s General Professional Proficiency target roughly sits.

C1: fluent without visibly searching

Spoken interaction. “I can express myself fluently and spontaneously without much obvious searching for expressions. I can use language flexibly and effectively for social and professional purposes. I can formulate ideas and opinions with precision and relate my contribution skilfully to those of other speakers.”

Spoken production. “I can present clear, detailed descriptions of complex subjects integrating sub-themes, developing particular points and rounding off with an appropriate conclusion.”

The phrase that separates C1 from B2 is without much obvious searching for expressions. A B2 speaker finds the word; a C1 speaker finds it without the listener noticing the search. The other C1 marker is relational: relating your contribution to what other speakers said. That is conversational skill, not vocabulary. You are no longer producing your points into the room, you are building on other people’s.

C2: the difficulty stops being visible

Spoken interaction. “I can take part effortlessly in any conversation or discussion and have a good familiarity with idiomatic expressions and colloquialisms. I can express myself fluently and convey finer shades of meaning precisely. If I do have a problem I can backtrack and restructure around the difficulty so smoothly that other people are hardly aware of it.”

Spoken production. “I can present a clear, smoothly-flowing description or argument in a style appropriate to the context and with an effective logical structure which helps the recipient to notice and remember significant points.”

Read the C2 interaction descriptor closely, because it is more forgiving than people assume. It does not say you never have a problem. It says that when you do, you route around it invisibly. C2 is not native-speaker perfection and the framework never claims it is. It is total repair ability plus precision of meaning. You still hit gaps; nobody hears them.

Why your own guess about your speaking level runs high

Here is the part that most CEFR guides leave out, and the reason this guide exists.

Almost every learner arrives at the CEFR through self-assessment. You read the grid, you find the row that sounds like you, and you announce a level. So the interesting question is not what the levels are. It is how good people are at placing themselves in them.

In 2024, Irina Stan published a study in the Open Journal of Modern Linguistics titled Self-Assessment versus Objective Proficiency Evaluation in English as a Foreign Language among Italian First-Year University Students. She had 1,090 first-year students at a large university in northern Italy rate themselves against the CEFR can-do descriptors, then compared each self-rating against the same student’s score on Pearson’s Versant English Placement Test.

The correlations, by skill:

Skill Self-assessment vs test (Pearson r)
Listening 0.68
Writing 0.58
Speaking 0.55
Reading 0.50

Stan’s own conclusion is positive, and it deserves stating fairly: all four correlations were statistically significant, all were in the range that self-assessment research typically reports, and she credits the CEFR’s can-do format for making self-rating work as well as it does. Specific, behavioral descriptors beat vague ones. The grid is doing its job.

But look at where speaking lands. It is third of four, and there is a systematic tilt underneath the number. The average student rated their own speaking more than half a level above where the placement test put it, on a scale whose steps are CEFR levels. Reading and speaking showed the largest gaps between self-rating and tested result, and more students overestimated their speaking than any skill except reading. The literature Stan reviews points the same way: an earlier study of 323 learners across eight languages found only a moderate correlation, r = 0.49, between self-assessed global speaking and a Defense Language Institute oral proficiency interview, and one 1996 placement study concluded that placement by self-assessment was “extremely unreliable.”

Why speaking in particular? Because it is the only skill you cannot observe yourself doing. When you read, you either follow the text or you do not, and the feedback is immediate and private. When you speak, you are busy speaking. You do not hear your own hesitations, your grammar simplifications, or the fact that the person opposite you quietly slowed down. Comprehension has a built-in error signal. Production does not.

There are two practical conclusions, and both are inversions of what people normally do.

Do not treat your self-rating as a fact. Treat it as a bracket of two adjacent levels. If you read the grid and land on B1, your real spoken interaction is somewhere in A2 to B1. That is not pessimism, it is the measured direction of the error.

Do not treat the gap as a personal failing. This is 1,090 motivated university students who had studied English for years, using the best-designed self-assessment instrument in the field, and speaking still came out third of four. Miscalling your own speaking level is the normal result, not evidence that you are worse than you thought.

How many words each level actually takes

Vocabulary is the one part of the CEFR you can put a number on, and it is worth doing, because “a broad lexical repertoire” is not an actionable target.

The CEFR’s own vocabulary range criterion is qualitative. At A1 a learner “has a basic vocabulary repertoire of isolated words and phrases related to particular concrete situations.” At B1 they have “a sufficient vocabulary to express him/herself with some circumlocutions on most topics pertinent to his/her everyday life such as family, hobbies and interests, work, travel and current events.” At C2 they have “a very good command of a very broad lexical repertoire including idiomatic expressions and colloquialisms.” Useful as description, useless as a goal.

James Milton of Swansea University put numbers to it in The development of vocabulary breadth across the CEFR levels, a chapter in the EUROSLA monograph on communicative proficiency and linguistic development. Two findings are worth carrying around.

First, the historical one. The original Council of Europe materials came with actual word lists. What is now B1 was called Threshold, or Niveau Seuil, and its word lists typically contained about 2,000 words. What is now A2 was called Waystage, and its lists ran to about 1,000 words. The lists were later dropped in favor of holistic descriptors, so that the framework could stretch across many languages, but they were never disowned.

Second, the measured one. Milton reports estimates from Cambridge exam candidates at every CEFR level, tested with the XLex vocabulary size test, which scores how many of a language’s most frequent 5,000 lemmatized words a learner knows.

CEFR level Words known, of the most frequent 5,000 (English)
A1 under 1,500
A2 1,500 to 2,500
B1 2,750 to 3,250
B2 3,250 to 3,750
C1 3,750 to 4,500
C2 4,500 to 5,000

Milton’s own summary of the pattern is blunt: something like 2,000 of the most frequent 5,000 words appears to be needed to reach A2, and roughly 3,000 to get past the elementary levels into any kind of independence. C1 and C2 are associated with near-complete recognition of the most frequent 5,000, which in turn implies a total vocabulary far larger. Drawing on Paul Nation’s coverage research, Milton notes that knowledge of the most frequent 5,000 words plus an overall vocabulary of perhaps 8,000 or 9,000 words is associated with the highest levels of fluency and understanding in English as a foreign language.

Three caveats keep this honest. These are English figures from English exams, and Milton’s own cross-language data suggests the totals differ by language, with French learners appearing to reach comparable levels on fewer words. These are recognition counts, not the words you can produce in speech, and the productive number is always smaller. And vocabulary is a necessary condition, not a sufficient one: knowing 3,000 words does not make you B1 if you have never assembled them out loud under time pressure. Frequency-first vocabulary is the cheapest leverage in language learning, which is a large part of what polyglots do differently, but it is leverage on a skill you still have to practice.

How long it takes to move up a level

The CEFR publishes no hours per level, and any guide that gives you a clean “A1 takes 90 hours” table is quoting a school’s course lengths, not the framework. But there is a defensible way to get an order of magnitude.

The Foreign Service Institute, the US State Department’s language school, publishes time estimates to reach General Professional Proficiency, which sits at roughly B2 to C1. For an English speaker learning Spanish, French, Italian, or Portuguese, that is about 600 to 750 hours of instruction from zero. For Mandarin Chinese, Japanese, Korean, or Arabic it is around 2,200 hours.

Spread across the four or five CEFR steps between nothing and B2, the easy-language figure works out to something on the order of a hundred-plus hours of contact per step, and the hard-language figure to roughly four times that. Two things follow.

The steps are not equal. A1 to A2 is a short climb; B2 to C1 is a long one, because each band above covers more ground than the one below it. Expect the pace to slow as you go up, and do not read the slowdown as a plateau.

And the hours only count if they are the right hours. FSI hours are intensive class time with heavy speaking. Six hundred hours of reading and flashcards will not produce a B2 speaker, because none of the speaking descriptors above describe anything you can do silently. Which brings us to the part that matters.

What actually moves you up a speaking level

Read all twelve descriptors above again and notice what they have in common. Every single one is a verb about doing something with another person in real time: interact, ask, answer, communicate, handle, deal with, enter unprepared, take an active part, account for, sustain, relate, backtrack, restructure.

Not one of them describes knowledge. There is no CEFR speaking descriptor that says “knows the subjunctive” or “has completed 400 lessons.” The framework was written this way on purpose, and it has a consequence most learners never draw: the only activity that reliably moves your spoken interaction up a level is spoken interaction. Everything else is preparation for it.

That reframes what to stop doing, which is usually more useful than a list of what to start.

Stop treating input as speaking practice. Reading and listening build the raw material and they raise your other CEFR rows. They do not move spoken interaction, because none of the descriptors can be satisfied while silent. This is the single largest misallocation in language learning.

Stop rehearsing only what you can already say. Every descriptor above has a boundary condition attached: A1 needs the other person to help, A2 cannot keep the conversation going, B2 can sustain a view under pushback. You move up by operating at the edge of your current level often enough that the edge moves, not by getting smoother inside it.

Stop waiting until your grammar is clean. The descriptors do not require accuracy. B1 explicitly allows circumlocution. C2 explicitly allows problems, provided you can route around them. Accuracy improves as a byproduct of volume, not as a prerequisite for starting. We made the fuller case for this in difficulty is not value in language learning.

Stop practicing only in one direction. Solo drills such as narration, shadowing, and recording yourself are excellent for spoken production and cost nothing, and how to practice speaking a language alone covers the five that work best. But spoken interaction has requirements solo practice structurally cannot meet: an unscripted question, a response to what you actually said, and a correction when you are wrong. The interaction row moves only when something on the other side is genuinely reacting.

Stop measuring in lessons and start measuring in minutes spoken. If you want a single metric that predicts CEFR speaking movement better than any other, it is cumulative time spent producing the language in a live exchange. It is also the metric almost nobody tracks.

How Mintza handles levels

Mintza is a voice conversation app where you talk with a bilingual AI teacher in real time, like a phone call, in fifteen languages: English, Spanish, Portuguese, French, Italian, German, Greek, Chinese, Russian, Turkish, Swedish, Arabic, Japanese, Korean, and Hebrew. Any pair, any direction. It is worth explaining how it treats levels, because it makes a deliberate choice that connects directly to the research above.

Mintza does not ask you to identify as a B1 or a C1. There is no CEFR label anywhere in the app. When you set up a language, it asks “How much do you know?” and offers four answers in plain words: Starter, meaning “I don’t know a single word.” Beginner, “I know some phrases. I speak a little.” Intermediate, “I get by speaking. I want to loosen up.” Advanced, “I can hold a fluent conversation.” The hint under the question is honest about why it is asking: “This helps your teacher start in the right place.”

That is the right question to ask a person whose self-rated speaking level, per Stan’s 1,090 students, tends to run more than half a level high. A four-way plain-language question is harder to get wrong than a six-way technical one, and the consequence of getting it slightly wrong is a first conversation pitched a little off, not a misdiagnosis you carry for a year. Your level is also saved per language, so you can be Advanced in French and Starter in German on the same account, which is the same principle the CEFR applies across skills applied across languages.

For orientation only, and this is our own approximate read rather than any official mapping from Mintza or the Council of Europe: Starter sits around Pre-A1 and A1, Beginner around A2, Intermediate around B1 and B2, Advanced at C1 and above.

What the app is actually built to produce is the thing the descriptors demand. It is open conversation, not a script, so you get the unprepared questions that B1 requires and the pushback that B2 requires. It corrects you inside the conversation instead of freezing it into a report afterwards. When a sentence collapses, it switches into the language you already speak, gets you moving, and brings you back, which turns the exact moment that ends most real conversations into a repair, the skill the C2 descriptor is built around. You also pick the regional accent you want to sound like, from British or Australian English to Madrid or Buenos Aires Spanish, and which accent should I learn covers how to choose.

You start with 20 free minutes, no card and no subscription required, and those free minutes never expire. Paid plans are monthly minute pools: Basic with 180 minutes, Plus with 360, and Pro with 600, cancel anytime, with current prices shown in the store. Minutes are the honest unit here, given that time spent speaking is the thing that moves the level.

The short version

The CEFR describes what you can do, not what you have studied, and it expects your five skills to sit at different levels. Read the six levels through their spoken interaction and spoken production descriptors and a clean progression appears: at A1 someone helps you, at A2 you transact but cannot sustain, at B1 you enter unprepared, at B2 the other person stops adjusting, at C1 you stop visibly searching, at C2 your problems become invisible.

Then hold your own self-rating loosely, because the research says speaking is where learners are most likely to place themselves too high, and that this happens to almost everyone. Bracket yourself across two levels instead of claiming one. Use vocabulary counts as a floor, not a finish line. And remember that every speaking descriptor in the framework is a verb performed with another person in real time, which means there is exactly one activity that moves those rows.

Mintza is available for iPhone, iPad and Mac and Android.