All articles
Behind the scenesOctober 2026·6 min read

How we score the placement test: a CEFR level for each skill

Your level comes from a ladder

Every task is tagged to a CEFR level (A1–C1) and a skill (reading, listening, grammar). Each level has 9 tasks, 3 per skill. We climb the ladder from A1 upward: to clear a level you need 6 of its 9 tasks right, and your level is the highest one you clear without a gap below it. For C1 you need 7 of 9.

Example
Each dot is one task
C1
2 / 9 · needs 7
B2
6 / 9 · gap below
B1
4 / 9 · not cleared
A2
7 / 9 · your level
A1
8 / 9 · cleared
ReadingListeningGrammarNot right

That's why the raw percentage alone can't name a level:

40% on the easy itemsA2

C1
B2
B1
A2
A1

Land your 40% entirely on the easy items and you're solidly A2.

The same 40%, scatteredA1

C1
B2
B1
A2
A1

Scatter the same 40% and miss the basics, and the ladder stops at A1.

Each skill gets its own level the same way, from its 3 tasks per level: 2 of 3 clears a level, and C1 needs all 3.

2 of 3 · A1–B2
3 of 3 · C1

Why one careless slip doesn't sink you

CEFR is cumulative: if you can do B2, you can almost certainly do A2. So when you clearly clear higher levels but stumble on a lower one, we treat that stumble as noise and the ladder keeps climbing instead of collapsing to the floor. This only applies up to B1: B2 and C1 always have to be earned outright.

Skipped tasks are treated the same way. If you leave most of a level unanswered but clear a level above it, we step over the gap instead of stopping there, and your result names the parts you didn't answer. Tasks you answered wrongly still count as wrong, and if nothing above was cleared, blanks count as wrong too.

Speaking is scored on its own scale

The optional speaking part doesn't run through the same ladder. You talk to an AI examiner; it scores your answer from 0 to 100 on the criteria a real examiner uses, and that score maps straight to a CEFR band:

85+C1
70+B2
55+B1
35+A2

Speaking is only rated when you said enough to judge. Otherwise it is marked "not rated" and your result rests on the other skills. When it is rated, it counts like any other skill (see below).

Reading your result

Your result card shows one large level and a short level for each skill, for example:

Example
Your level
B1
48%weighted accuracy
Reading~B2
Listening~A2
Grammar~B1

The large level is never more than one step above your weakest skill

In the example, listening is A2, so the large level can be at most B1, even if reading points higher. If you took the speaking part, it counts as a skill here too.

Reading ~B2Listening ~A2Grammar ~B1weakest + 1 step →max B1

The percentage is the weighted accuracy

Under the large level you see a percentage: how many tasks you got right, where harder tasks count more.

"~B1" means "about B1"

Your grammar answers fit B1 best. It is an estimate, not an exact grade, because each skill has only 15 tasks.

Built to be fair

Three things keep the result honest:

01

Same number of tasks per level

Every level has 3 tasks in each skill, so no level depends on one lucky guess.

02

No answer patterns

The correct answers are spread evenly across the options, so picking the same letter every time gets you nowhere.

03

Harder texts at higher levels

Texts and recordings get longer and more complex from level to level.

Treat the result as an estimate. You can retake the test any time; a retake opens in a new tab, so your current result stays on screen. Use the skill breakdown to aim your prep where it will move you fastest.