CLB is not a CELPIP thing. The Canadian Language Benchmarks are Canada's own scale for describing what a person can do in English, and CELPIP is one of several tests calibrated against it — the scale is "calibrated against the Canadian Language Benchmark (CLB) levels, as well as the Common European Framework". The test does not define the benchmark. It reports against it.
That is also why your result arrives as four separate levels, one per skill. The scale describes abilities, not people. There is no such thing as "a CLB 9 person" — there is a CLB 9 in Listening, and a separate answer for Reading, and two more after that.
Paragon, the test's owner, gives the range:
"CELPIP test scores have been aligned to the Canadian Language Benchmarks (CLB) Levels 3 to 12."
The failure
You get four levels back. You average them, or quote the highest, or round the report into a single sentence — "I got a nine" — and plan from there.
Because programs read the four levels separately, your weakest skill is usually the one that decides. A 9, 9, 9 and a 6 is not "mostly nine". Where a requirement is written as a minimum in each of the four skills, that result is a six, and the three nines change nothing. The average is arithmetic. The requirement is a floor.
The mapping, with the edges handled honestly
Most of the chart is boring, and that is the good news: through the middle, CELPIP N is CLB N. The two ends are not the same kind of fact, and do not come from the same source.
| CELPIP Level | CLB level | Who publishes this row |
|---|---|---|
| 12 | 12 | Paragon guidebook |
| 11 | 11 | Paragon guidebook |
| 10 | 10 | IRCC chart |
| 9 | 9 | IRCC chart |
| 8 | 8 | IRCC chart |
| 7 | 7 | IRCC chart |
| 6 | 6 | IRCC chart |
| 5 | 5 | IRCC chart |
| 4 | 4 | IRCC chart |
| 3 | 3 | Paragon guidebook |
| 2 | "1, 2" — not one-to-one | Paragon chart |
| 1 | none — "Insufficient information to assess" | Paragon chart |
| 0 | none — "Insufficient information to assess" | Paragon chart |
Read the third column — it is the part other pages drop.
IRCC's own equivalency chart covers CLB 4 through 10, publishing the same 1:1 row for Reading, Writing, Listening and Speaking alike. That is the range where the government has published a number you can check against it.
CELPIP 11 = CLB 11 and CELPIP 12 = CLB 12 come from Paragon, not from IRCC — as does 3 = 3. That is the test owner's own statement of alignment: a real source, but a different one.
The bottom rows are not a mapping at all. CELPIP 2 corresponds to CLB "1, 2" — two levels, not one — so there is no clean equivalent to quote. Levels 0 and 1 carry "Insufficient information to assess" and map to no CLB level at all. A page printing "CELPIP 1 = CLB 1" has invented it.
For the CLB level a given program asks for, see Which CLB you actually need, by program — those requirements belong next to their IRCC citations, not restated here.
The Average Score trap
Your score report shows an Average Score. It is the most misread number there, and Paragon says exactly what it is worth:
"Average Score is not equivalent to any Canadian Language Benchmark (CLB) level"
This is where "not one number" stops being a philosophical point. People average their four levels and plan around a figure that does not exist as a benchmark — it is not on the CLB scale, and it cannot be compared to a CLB requirement at all. An average of 8.25 is not a CLB anything. The four levels underneath it are the result.
One more thing on the report: the old "M" score was replaced in the charts "with the 0-2 score" in December 2023. References to M are still all over older forum threads and prep sites. If you find one, you are reading pre-2024 material — worth knowing about the rest of that page.
Which skill to work on first
This is a question with an answer, and the answer is usually the lowest one, not the one you enjoy practising. If your levels come back uneven, the lowest is the one carrying weight — the others are already doing what they can.
It is not a satisfying answer, because the lowest skill is normally the one you avoid: Speaking if you dislike hearing yourself, Writing if you dislike being marked. Reading practice when Reading is already your best skill feels productive — you get most of them right, which is exactly why it changes nothing.
Two things make this concrete. Speaking and Writing are where a response giving insufficient information to assess scores 0 or 1 — a mechanical failure, not a language one. And Listening and Reading each report a raw score out of 38, so a weak section shows up as countable questions rather than a feeling.
What is not published
Two gaps, stated straight, because most pages fill them with confidence.
No official CELPIP-to-CEFR chart was found. The overview page says the scale is calibrated against the Common European Framework as well as CLB — but a stated calibration is not a conversion table, and none appeared on celpip.ca. If you see one, check whose it is.
No official raw-score cutoff is published, and this page prints none. CELPIP labels its own raw-score table "Approximate". The hard cutoff tables elsewhere are reverse-engineered by third parties and they contradict each other — different sites assign the same level to different raw ranges. Publishing one as fact is publishing a guess with a decimal point on it.
What actually costs you
- Quoting one number when a program reads four — the average, the highest, or an "overall" level that was never issued.
- Treating the Average Score as a CLB level. Paragon says in writing it is not one.
- Assuming CELPIP 11 or 12 = CLB 11 or 12 is an IRCC-published row. It is Paragon's.
- Quoting a CLB equivalent for CELPIP 0, 1 or 2. There isn't a clean one, and for 0 and 1, none.
- Trusting a prep site's raw-score cutoff, then treating a practice result as a predicted level.
- Practising your strongest skill because it feels better than practising your weakest.
- Reading advice written before December 2023 without noticing — the "M" score is the giveaway.
Requirements change, and IRCC's own page is the only place they are authoritative. Check it before you act.
Try one
[Try a marked task →] — one full task, marked, no card required.