The short answer

What the average IQ is, and why it is 100 at every age

The average IQ score is 100. Scores between 90 and 109 are the average band and hold about half the population. Widen it to 85 to 115 and you have about two thirds of everyone. That is true for a nine-year-old, a nineteen-year-old and a seventy-nine-year-old, because each is compared with their own age group.

This is the part most people miss. An IQ is not a count of anything. It is a rank turned into a number: your performance is placed against a norm sample of people your age, converted to a standard score, and published on a scale built to average 100 with a standard deviation of 15.3 Change the age group and you change the yardstick.

So "what is the average IQ for a 12-year-old" has a slightly frustrating answer: 100, because the test was designed to make it 100. The useful question underneath it is usually one of two others. Either "is my child's score normal", which the band above answers, or "does raw thinking ability change with age", which it certainly does, and which we measured on 91,695 of our own sittings further down this page.

The bands are the same at every age as well. These four cover the language people actually use, and the full IQ scale chart has all seven levels with percentiles.

What each IQ range is called, at any age
ScoreUsually calledShare of people
130 and aboveVery Superior, and the usual gifted threshold2.2%
120 to 129Superior, clearly above average6.7%
110 to 119High Average16.1%
90 to 109Average, the normal band50%
Below 90Low Average and below25%

The instrument changes with age too. Children from two and a half to seven are usually given the WPPSI, six to sixteen the WISC, and sixteen upward the WAIS.12 Each is normed separately, which is why a score can shift by a few points when a teenager crosses from one to the next without anything about them having changed.

The table

Average IQ by age, from 4 to 80 and over

The first three columns are the same at every age, and that is the point. The last two are ours, and they are where the differences actually show up.

Average IQ, normal band and test used at each age, with our own measured figures
Age Average IQ Normal band Below the strongest band, on our test Median minutes taken
Under 13 100 90 to 109 -8.3 points 6.4
13 to 15 100 90 to 109 -3.8 points 7.6
16 to 17 100 90 to 109 -2.2 points 8.1
18 to 24 100 90 to 109 -0.8 points 8.9
25 to 34 100 90 to 109 strongest band 9.5
35 to 44 100 90 to 109 -2 points 9.9
45 to 54 100 90 to 109 -2.9 points 10.4
55 to 64 100 90 to 109 -5 points 11.4
65 and over 100 90 to 109 -7.8 points 11.3

Every single year from 4 to 19

Most people arrive here looking for one particular age. Here is each one on its own row, with the test a school or clinic would normally use at that age.

Average IQ, normal band and test used at each single year of age from 4 to 19
AgeAverage IQNormal bandTest normally used
4 100 90 to 109 WPPSI-IV, ages 2 years 6 months to 7 years 7 months
5 100 90 to 109 WPPSI-IV, ages 2 years 6 months to 7 years 7 months
6 100 90 to 109 WPPSI-IV, ages 2 years 6 months to 7 years 7 months
7 100 90 to 109 WPPSI-IV, ages 2 years 6 months to 7 years 7 months
8 100 90 to 109 WISC-V, ages 6 to 16
9 100 90 to 109 WISC-V, ages 6 to 16
10 100 90 to 109 WISC-V, ages 6 to 16
11 100 90 to 109 WISC-V, ages 6 to 16
12 100 90 to 109 WISC-V, ages 6 to 16
13 100 90 to 109 WISC-V, ages 6 to 16
14 100 90 to 109 WISC-V, ages 6 to 16
15 100 90 to 109 WISC-V, ages 6 to 16
16 100 90 to 109 WISC-V, ages 6 to 16
17 100 90 to 109 WAIS-5, ages 16 to 90
18 100 90 to 109 WAIS-5, ages 16 to 90
19 100 90 to 109 WAIS-5, ages 16 to 90

Read the fourth column of the band table carefully. It is not an IQ by age. Our test is deliberately not age-normed: everyone answers the same items and is scored against one reference sample, which is exactly what lets it show the gradient a clinical test hides. The figures are gaps from the strongest band, measured only on sittings that took six to fifteen minutes so that time spent is held roughly constant.

One age at a time

What is the average IQ for a 4, 8, 10, 12, 15 or 18 year old

These are the ages people actually search for, and the answer is 100 for all of them. What changes is which test is used, how much the score moves between sittings, and how much weight it is safe to put on it.

Average IQ for a 4-year-old

100, with 90 to 109 counting as the normal band. Assessed with the WPPSI-IV, for ages 2 years 6 months to 7 years 7 months.

At this age a score moves a lot between sittings, and attention on the day counts for more than it will later. Treat a preschool result as a rough marker, not a forecast.

Average IQ for a 8-year-old

100, with 90 to 109 counting as the normal band. Assessed with the WISC-V, for ages 6 to 16.

By eight, scores start to become more stable from year to year, but a single result still carries a band of about five points either side.

Average IQ for a 10-year-old

100, with 90 to 109 counting as the normal band. Assessed with the WISC-V, for ages 6 to 16.

A ten-year-old is compared with other ten-year-olds, so 100 is average and 90 to 109 is the normal band. A raw jump in ability over the year can leave the IQ unchanged, because everyone else grew too.

Average IQ for a 12-year-old

100, with 90 to 109 counting as the normal band. Assessed with the WISC-V, for ages 6 to 16.

This is the most searched age of all. The answer is still 100. Twelve-year-olds are in a fast growth phase, so a score taken now and a score taken at fifteen can differ without anything being wrong.

Average IQ for a 15-year-old

100, with 90 to 109 counting as the normal band. Assessed with the WISC-V to 16, then WAIS-5.

Around this age a test-taker crosses from the childrens scale to the adult one. The two are normed separately, so a small step in the number at the switch is a change of instrument, not of mind.

Average IQ for a 18-year-old

100, with 90 to 109 counting as the normal band. Assessed with the WAIS-5, for ages 16 to 90.

From here the same adult scale runs for the rest of life, still age-banded. Raw reasoning is close to its peak, which is why this is the age group most likely to mistake a good day for a ceiling.

For an age not listed, the answer is the same 100. Use the lookup at the top of the page for the test used and our own figures at that age.

Our own data

What 91,695 sittings show about age and reasoning

Because our test is not age-normed, it can answer the question a clinical test is built to remove: how does raw performance on the same items differ across ages? These are 91,695 completed United States sittings taken between 23 June and 2 September 2026, grouped by the age band each person selected before starting the test.

  • 91,695US sittings, by age band
  • 25 to 34strongest age band
  • 8.3 ptsgap to the under-13s
  • 7.8 ptsgap to the over-65s

Raw reasoning peaks between 25 and 34, and the curve is symmetrical

Performance climbs through the teens, peaks in the twenties and early thirties, then falls away at almost the same rate it rose. The under-13s sit 8.3 points below the peak band and the over-65s 7.8 points below it. That inverted U is the fluid reasoning curve that cognitive ageing research describes, and it is the reason the clinical scales bother with age norms at all.4

How far each age band falls below the strongest band on our test Scores rise from 8.3 points below the peak among under-13s to the peak at 25 to 34, then fall away again to 7.8 points below the peak at 65 and over. Sittings are limited to those taking six to fifteen minutes, so time spent is held roughly constant. 0 -2 -4 -6 -8 Under 13: 8.3 points below the peak, 6,532 sittings -8.3 Under 1313 to 15: 3.8 points below the peak, 12,047 sittings 13-1516 to 17: 2.2 points below the peak, 7,144 sittings 16-1718 to 24: 0.8 points below the peak, 9,208 sittings 18-2425 to 34: strongest band, 6,408 sittings peak 25-3435 to 44: 2 points below the peak, 3,866 sittings 35-4445 to 54: 2.9 points below the peak, 2,552 sittings 45-5455 to 64: 5 points below the peak, 1,579 sittings 55-6465 and over: 7.8 points below the peak, 1,441 sittings -7.8 65+ age band points below the strongest band
Points below the strongest age band, 50,777 US sittings that took six to fifteen minutes. Hover a point for its sample size.

The time control matters. Under-13s and over-65s rush more than anyone else: 19.6% and 13.3% of their sittings finish in under three minutes, against 5.8% at 25 to 34. Without that control the gap to the over-65s looks like 12.4 points. Holding time roughly constant cuts it to 7.8, so about a third of the apparent decline in the raw numbers was people not giving the test their time, and two thirds was real.

The same score is a different rank at every age

This is what an age-normed test is correcting for, made visible. Take one fixed raw score and ask where it lands inside each age band. A score of 120 sits near the middle of the under-13s and near the middle of the over-65s, but it is in the bottom fifth of the 25 to 34 band. The performance is identical. The company it keeps is not.

Percentile rank of four fixed raw scores within each age band
Age band Rank of a 100 Rank of a 110 Rank of a 120 Rank of a 130
Under 13 17th 24.1th 41.4th 75.8th
13 to 15 9.3th 13.5th 26.4th 64.3th
16 to 17 8.7th 11.7th 22.4th 59.8th
18 to 24 8.4th 10.6th 18.5th 52th
25 to 34 7.9th 9.7th 16.1th 46.9th
35 to 44 8.2th 11.3th 20th 52.6th
45 to 54 8.1th 11.8th 21.9th 57.4th
55 to 64 10.5th 15.3th 27.5th 62.2th
65 and over 23.9th 28.3th 41.6th 72.3th

Read a row across and you have that age band's own scale. Read a column down and you have the reason a twelve-year-old and a thirty-year-old with the same answers do not receive the same IQ. On a clinical test the correction is applied for you and you never see it. Here it is.

How to read these numbers

Sittings are self-selected, unsupervised and scored on a 33-item form built from the public ICAR item bank, against an online research norm rather than a census-matched one.10 Its absolute level therefore runs high, which is why nothing here is published as an average IQ by age. Every figure above is either a gap from the peak band or a rank inside a band, and both are unaffected by that inflation. Ages are self-reported at the start of the test. We use United States sittings only: the worldwide curve peaks at 13 to 15, which is a fact about who takes the test in each country rather than about cognition. Method notes are on the methodology page.

For parents

Reading a child's IQ score without over-reading it

Most people who look up an average by age are holding a number for one particular child. Four things are worth knowing before that number does any work.

It is a band, not a point. A well-administered childrens test reports a confidence interval, usually about five points either side. A 118 and a 123 are the same result wearing different clothes, and a report that gives a single integer with no interval is telling you less than it appears to.

It moves, and moves most when they are youngest. Scores taken before school age are unstable enough that they are mainly used to investigate a specific concern. Stability climbs through childhood; by the mid-teens it is high. The Scottish Mental Survey follow-up found scores at eleven still strongly related to scores from the same people at seventy-seven, which is about as good as psychological measurement gets, though it describes a group rather than a promise about one child.7

Practice moves it. A second sitting of a similar test lifts scores by roughly a quarter of a standard deviation, and more when the gap between sittings is short.9 If a child has been coached for an entry test, some of the gain is the coaching.

Check what was actually administered. A full assessment is one-to-one with a psychologist and takes an hour or more: the WPPSI up to about seven, the WISC from six to sixteen, the WAIS after that.12 A school-wide group test, an entrance exam or an online quiz is a different kind of measurement with a wider error band, and the report should say which one it was.

Below 100 is not a problem to solve. Half of all children score below 100, because that is what an average means. The band from 90 to 109 is squarely normal, and a score in the eighties still describes a child who will manage school and life perfectly well. A score is worth acting on when it comes with real difficulty in the classroom or at home, and then the thing to act on is the difficulty, not the number.

It says nothing about effort, kindness or how school will go. A cognitive score predicts academic performance at the group level and predicts very little about one child's week. Teachers, sleep, interest and the subject they happen to love will do more.

After 60

What happens to IQ in later life

Two abilities move in opposite directions, and the single IQ number hides the fact that they are doing so.

Fluid reasoning, the ability to solve a problem you have never seen before, peaks in early adulthood and declines slowly from around the late twenties. Crystallised ability, the store of vocabulary, facts and practised judgement, keeps growing well into the sixties and holds up long after.64 Our own curve measures the first of those, which is why it falls; a vocabulary test on the same people would slope the other way.

Because the scale is age-normed, an older adult and a younger adult with the same raw performance do not receive the same IQ. The older adult's is higher, and legitimately so: they are being compared with their own age group. The rank table above is that correction with a number on it, measured rather than asserted.

The practical reading is that a stable IQ across decades is the normal outcome, not a reprieve. A sharp unexplained fall in everyday functioning is a reason to see a clinician, and not something an online score can assess.

Honest limits

What an average IQ cannot tell you

An average describes a group. It cannot tell you where one person sits, and the spread inside every age band is far wider than the differences between the bands. On our own data the gap between the strongest and weakest age band is about eight points; the gap between the top and bottom quarter within a single band is several times that.

The number also drifts with history. Measured performance rose roughly three points a decade through the twentieth century, the Flynn effect, which is why tests are re-normed and why a score from a 1995 test and one from a 2025 test are read against different crowds.8 A parent comparing their own childhood score with their child's is comparing two different yardsticks.

And an age-normed average says nothing about ability in absolute terms. A society could get markedly better or worse at reasoning and every published average would still read 100. That is a design choice, not a finding, and it is the single most useful thing to keep in mind when reading any table of average IQ by age, including this one.

The average is useful for

  • Deciding whether one score is inside the normal band for that age.
  • Understanding why a score changed when the test edition changed.
  • Seeing how much of a difference between two people is inside the error band.

It is not useful for

  • Comparing a child with a parent, or a 1995 score with a 2025 one.
  • Ranking age groups against each other on the IQ number itself.
  • Predicting one person's school year, job or character.

Questions

Average IQ and age: common questions

Does IQ go down as you get older?

The IQ number usually does not, because it is re-centred on your own age group every time. What declines is raw speed and novel problem solving, from roughly the late twenties onward. Vocabulary, knowledge and judgement hold up or keep improving into the sixties, so an older adult can score the same 105 while solving problems in a different way.

Is there an average IQ by gender?

Test publishers balance their norm samples so that men and women average the same overall score, and large reviews find no meaningful difference in general ability. Small differences appear on some sub-tests in both directions, and the spread of scores is slightly wider among men. For a single person the sex of the test-taker tells you nothing useful about the result.

What is the average IQ in the United States?

By construction it is 100, because the American norm sample is what the scale is centred on. Country comparison tables that put the United States at 97 or 98 come from a different exercise entirely, one that stitches together separate studies of varying quality, and they are contested. Use the age-normed 100 for reading an individual score.

My child scored 120 at age 7 and 110 at age 11. Did their IQ drop?

Probably not. A ten-point move across four years is inside the range you would expect from measurement error, a different test edition and normal development. Childhood scores become more stable with age but are never fixed. Look at the confidence band the report prints, and at whether school and home life changed, before reading anything into the number.

My 10-year-old scored 94. Should I be worried?

No. A 94 is inside the normal band, and half of all children score below 100 by definition. Scores in the eighties and nineties describe children who cope with school perfectly well. What is worth attention is a specific difficulty, in reading or attention or a subject, and that is assessed by looking at the difficulty rather than at the composite number.

What is a good IQ for a 12-year-old?

The same as for anyone else: 90 to 109 is the average band, 110 to 119 is above average and 130 and over is the usual gifted threshold. There is no separate childrens scale of good and bad. If you are weighing a gifted programme, most use 130, and a few use 125 or a percentile cut instead.

Do IQ scores rise across generations?

Yes, and it is the reason tests are re-normed. Measured performance rose about three points a decade through the twentieth century, the Flynn effect, with the rate varying by domain and slowing recently. A score from a 1990s test and a score from a 2020s test are read against different comparison groups even when the number is the same.

At what age is IQ most stable?

Stability climbs steadily through childhood and is high from the mid-teens onward. The Scottish Mental Survey follow-up found scores taken at age 11 still correlated strongly with scores from the same people at 77, which is a remarkable result for any psychological measure, though it describes a group and not a promise about one person.

Can a toddler be given an IQ test?

There are scales normed from about two and a half years, but results at that age are unstable and are mostly used to investigate a specific developmental concern rather than to rank a child. A number taken before school age should never be treated as a forecast.

Or sit a scored test, free and with no signup

Michael Hodge, Founder & Assessment Director. Bachelor of Science (Psychology), University of Wollongong. with the age-band analysis, the rank table and the per-age answers.

Sources (10)
  1. Pearson Assessments. Wechsler Intelligence Scale for Children, Fifth Edition (WISC-V)
  2. Pearson Assessments. Wechsler Preschool and Primary Scale of Intelligence, Fourth Edition (WPPSI-IV)
  3. Pearson Assessments. Wechsler Adult Intelligence Scale, Fourth Edition (WAIS-IV)
  4. Salthouse, Journal of the International Neuropsychological Society (2010). Selective review of cognitive aging
  5. Salthouse, Neurobiology of Aging (2009). When does age-related cognitive decline begin?
  6. Horn and Cattell, Journal of Educational Psychology (1966). Refinement and test of the theory of fluid and crystallized general intelligences
  7. Deary, Whalley, Lemmon, Crawford and Starr, Intelligence (2000). The stability of individual differences in mental ability from childhood to old age: follow-up of the 1932 Scottish Mental Survey
  8. Pietschnig and Voracek, Perspectives on Psychological Science (2015). One century of global IQ gains: a formal meta-analysis of the Flynn effect
  9. Scharfen, Peters and Holling, Intelligence (2018). Retest effects in cognitive ability tests: a meta-analysis
  10. Condon and Revelle, Intelligence (2014). The International Cognitive Ability Resource: development and initial validation of a public-domain measure