TOEFL's New 1 to 6 Scale Is Hardest on Speaking

TOEFL changed the shape of its score report in January 2026. The band scale is the headline, and the conversion table underneath it is the story.

Speaking demands a higher raw score than any other section for the same band, across most of the scale.

SpeakShark is free for the section the new scale is least forgiving about, three AI conversation sessions a day with no card.

In this guide: what changed · the conversion table · speaking · equal scores · the average · the old total · regulators · limits · method · FAQ

Key takeaways

From 21 January 2026, TOEFL iBT reports on a 1 to 6 scale in half points.
The overall score is the average of the four sections, rounded to the nearest half band.
Speaking needs the highest raw score of any section for bands 2 through 5.
A raw 24 is band 5 in reading, listening and writing, and 4.5 in speaking.
The old 0 to 120 total continues for two years, as a midpoint rather than your total.

What changed on 21 January 2026

ETS states that as of 21 January 2026, TOEFL iBT score reports feature a new 1 to 6 scale, in half point increments, with a more intuitive alignment to the Common European Framework of Reference for Languages.

Three things follow from its own notes:

You receive four section scores and an overall score on the 1 to 6 scale.
The overall is the average of all four sections, rounded to the nearest half band. ETS gives the example that an average of 5.25 produces 5.5.
For a two year transition period, you also receive a comparable overall score on the 0 to 120 scale, representing the midpoint of the corresponding total range.

The alignment to CEFR is the stated motivation, and it puts TOEFL on the same kind of instrument as everything else in this series. CEFR levels for speaking explains what those levels describe, and IELTS for UKVI by CEFR level shows the comparable mapping for the test TOEFL is most often set against.

The conversion table

ETS publishes the comparison for tests taken before 21 January 2026. This is the whole table, and it repays reading down the columns rather than across the rows.

Band	Reading	Listening	Speaking	Writing	Total
6	29 to 30	28 to 30	28 to 30	29 to 30	114+
5.5	27 to 28	26 to 27	27	27 to 28	107+
5	24 to 26	22 to 25	25 to 26	24 to 26	95+
4.5	22 to 23	20 to 21	23 to 24	21 to 23	86+
4	18 to 21	17 to 19	20 to 22	17 to 20	72+
3.5	12 to 17	13 to 16	18 to 19	15 to 16	58+
3	6 to 11	9 to 12	16 to 17	13 to 14	44+
2.5	4 to 5	6 to 8	13 to 15	11 to 12	34+
2	3	4 to 5	10 to 12	7 to 10	24+
1.5	2	2 to 3	5 to 9	3 to 6	12+
1	0 to 1	0 to 1	0 to 4	0 to 2	0+

Section scores are out of 30 and the total is out of 120.

Speaking is the strictest column

Read the floors rather than the ranges and the pattern is unmistakable.

Band	Lowest reading	Lowest listening	Lowest speaking	Lowest writing
5	24	22	25	24
4.5	22	20	23	21
4	18	17	20	17
3.5	12	13	18	15
3	6	9	16	13
2.5	4	6	13	11
2	3	4	10	7

For every band from 2 up to 5, speaking requires the highest raw score of the four.

The gap is widest in the middle. At band 3, reading needs 6 out of 30 and speaking needs 16. That is ten raw points of difference for the same reported band.

Only at the very top does the pattern break: at band 6, speaking starts at 28 while reading and writing start at 29.

One more detail worth pausing on. At band 5.5, speaking is a single value: 27. Every other section has a two point window there. In speaking, 26 is a 5 and 28 is a 6, and the half band in between is one number wide.

We are not going to speculate about why the scale is built this way, because ETS does not publish the reasoning on this page. What matters to a candidate is the arithmetic, and the arithmetic is one sided.

The same raw score is not the same band

Here is the practical consequence, and it is the thing most likely to surprise somebody reading a converted report.

Take a candidate who scored a flat 24 out of 30 in all four sections before the change.

Section	Raw	Band
Reading	24	5
Listening	24	5
Writing	24	5
Speaking	24	4.5

Three fives and a four and a half from four identical numbers. The average is 4.875, which rounds to an overall of 5.

The overall survives. The section score does not, and a section score is exactly what a regulator with a floor under speaking will be reading.

The overall is a plain average

Because the overall is the arithmetic mean of four sections rounded to the nearest half band, each section carries a quarter of the weight and nothing is discounted.

That is a different instrument from a rule that puts a floor under every skill, like the HCPC requirement of nothing below 6.5, and different again from Ireland's nursing rule, which allows one component to sit lower and lets the candidate choose which, covered in Irish nursing English.

An average is forgiving of one weak section and unforgiving of two. Half a band lost in speaking costs an eighth of a band overall, which usually disappears in the rounding. A band lost in speaking and half a band in writing does not.

The 0 to 120 score is now a midpoint

The transition detail deserves more attention than it usually gets.

For two years, reports also carry a comparable overall score on the 0 to 120 scale representing the midpoint of the corresponding total range.

That is not your old total. It is the middle of the range your new band maps to. Two candidates who would have scored 96 and 105 under the old scale both land on band 5, and both receive the same comparable number.

For most readers that changes nothing. For an institution comparing applicants finely on a 120 point scale, it removes resolution that used to be there, and a candidate who was strong within their band no longer has a way to show it.

Transition windows like this are normal when a test is replaced. LanguageCert's ESOL SELT product was replaced on 6 July 2025 with a two year window for existing certificates, which we cover in LanguageCert Academic against General SELT and in English certificates that expire.

What this does to published requirements

Every body that named a TOEFL number has to restate it, and they do not all restate it the same way.

The HCPC is a worked example. Before 21 January 2026 it asked for 100 out of 120 for most professions, with no section minimums published on that page. From that date it asks for reading 4.5, listening 5, writing 4.5, speaking 5, and overall 5.

Set the two against the conversion table and they are not the same requirement translated.

The old rule was a single total with no floor under any section.
The new rule is four floors plus an overall, and the overall band of 5 maps to a total of 95 and above rather than 100.

So the headline number came down slightly and four new constraints appeared underneath it. The tightest of them is speaking 5, which on the old scale meant 25 out of 30.

That is one regulator. The general lesson is that a remembered TOEFL number is now unreliable, and the only safe move is to read your target body's current wording. Other regulators handle it differently again: the GMC works in IELTS and OET, covered in the GMC requirement for doctors, and the HCPC's country exemptions can remove the question entirely, covered in the 43 qualifying countries.

If speaking is the section carrying the strictest conversion, it is also the one that does not improve by reading quietly. Role play scenarios and a daily speaking partner are the practical routes, the TOEFL speaking section format covers the tasks themselves, and testing your speaking level free gives you a baseline. For how examiners mark spoken English in detail, Cambridge speaking criteria by level works from real scored candidates, and what C2 actually means sets the top of the scale.

What we could not verify

We did not verify the CEFR alignment itself. ETS describes the new scale as more intuitively aligned to CEFR. We have not seen a published band to level mapping and have not assumed one.

We did not establish why speaking converts more strictly. ETS publishes the table and not the reasoning, and we have deliberately left the question open rather than inventing an explanation.

We did not check how universities are treating the change. Recognition and admissions policy is set by each institution.

We did not verify the HCPC figures against ETS. The old and new HCPC thresholds are as the HCPC publishes them; the conversion table is as ETS publishes it. Setting them side by side is our arithmetic, not either body's statement.

The transition window has an end. ETS describes two years from January 2026, so the comparable 0 to 120 score will stop appearing.

This is not admissions, registration or immigration advice. It is a reading of published scoring documents, linked below.

How we researched this guide

We went looking for a regulator's rule and found a scale change underneath it. The HCPC page carries two sets of TOEFL minimums separated by a date, which is the sort of thing that only makes sense if the test moved, so we went to ETS to find out what moved.

The finding is in the table rather than the announcement. The announcement says the scale is now 1 to 6 and aligned to CEFR, which sounds like presentation. The conversion table says the four sections map to bands on four different curves, which is substance.

Reading the columns instead of the rows is what surfaced it. Row by row, the table looks like a tidy set of ranges. Column by column, speaking sits above every other section at almost every band, by as much as ten raw points in the middle of the scale.

The flat 24 example came last, as a check. If the claim is real it should produce a concrete case where identical performance reports unevenly, and it does.

We stopped short of explaining the cause. A scaled score can be built many ways and ETS does not publish the construction here, so the honest position is that the map is one sided and we do not know why.

Practise speaking, from SpeakShark

SpeakShark is an AI English speaking practice app, and this is the clearest case in the series for practising one section on purpose.

The conversion table says the same effort spread evenly across four skills does not report evenly. Speaking is where raw points are most expensive, which means it is also where they buy the most.

You talk, the AI answers what you actually said, and you get speaking feedback while the conversation is still running. The free tier gives basic feedback; the detailed pronunciation and grammar breakdown is on Premium.

Being straight about the limits: a free session runs five minutes with four turns, and we issue no certificates. We are not affiliated with ETS and we do not reproduce TOEFL task types.

Start free with three sessions a day and no card. Paid sessions run ten minutes with unlimited turns. Limits are on the pricing page, and how it works walks through a session. Sign up here if speaking is the column deciding your band.

We are a speaking improvement tool. We are not an exam preparation provider, and we are not affiliated with Cambridge English, the IELTS partners, Pearson or any exam board. For test format, booking and official practice material, go to the exam body directly. Use SpeakShark to make your spoken English stronger, and use official material to learn the test.

Site: speakshark.com
Support: speaksharksupport@gmail.com
Free tier: 3 AI conversation sessions per day, no card

Sources

ETS, understand your TOEFL iBT scores: the 21 January 2026 change to a 1 to 6 scale in half point increments, the CEFR alignment statement, the four section scores plus overall, the overall as the average of all four sections rounded to the nearest half band with the 5.25 to 5.5 worked example, the two year transition providing a comparable 0 to 120 overall representing the midpoint of the corresponding total range, and the full comparison table for tests taken before 21 January 2026
HCPC, certificate of English language proficiency: the 100 out of 120 requirement for tests before 21 January 2026 and the section minimums from that date, used as a worked example of a regulator restating a requirement
Council of Europe, CEFR Table 1, common reference levels global scale: the framework the new scale is described as aligning to

FAQ

What changed in TOEFL scoring in January 2026?

As of 21 January 2026, TOEFL iBT score reports use a 1 to 6 scale in half point increments, which ETS describes as a more intuitive alignment to the Common European Framework of Reference. You receive four section scores and an overall score on that scale, and for a two year transition period you also receive a comparable overall score on the old 0 to 120 scale.

How is the new overall TOEFL score calculated?

It is the average of all four section scores, rounded to the nearest half band. ETS gives the worked example that an average of 5.25 produces an overall of 5.5. That makes it a plain arithmetic mean, so one weak section pulls the overall down by a quarter of its own shortfall with nothing to offset it but the other three.

Does the old 0 to 120 score still appear?

For a two year transition period after January 2026, yes. ETS states you will also receive a comparable overall score on the 0 to 120 scale, representing the midpoint of the corresponding total range. That is a midpoint rather than your actual old style total, which is a meaningful difference if an institution is reading it closely.

Which TOEFL section is hardest to band?

Speaking, across most of the scale. On the published conversion table, band 3 needs 16 out of 30 in speaking against 6 out of 30 in reading. Band 5 needs 25 in speaking against 22 in listening. Only at the very top does speaking stop being the section demanding the highest raw score.

Do identical section scores give identical bands?

No, and this catches people out. A raw 24 out of 30 is band 5 in reading, listening and writing, and band 4.5 in speaking. Each section has its own conversion, so a score report that looked even under the old scale can come back uneven under the new one.

Does the new scale change what regulators require?

It forces them to republish. The HCPC, for example, kept 100 out of 120 for tests taken before 21 January 2026 and set separate section minimums from that date, including speaking 5 and overall 5. Anyone with a target score should read their target body's current wording rather than a remembered number.