IQ percentile: what your score means

Short answer: An IQ percentile is the share of people in a test's norm group who scored at or below your score. It is not the share of questions you got right. On the standard IQ scale, with an average of 100 and a standard deviation of 15, the normal curve puts 100 at the 50th percentile, 115 at the 84.1st, 120 at the 90.9th and 130 at the 97.7th. Every score carries error, so read your percentile as a range too.

What an IQ percentile means

A percentile rank compares you with other people. The Leaders Project at Teachers College, Columbia University defines it as the share of scores in the norm sample that were equal to or lower than yours. It warns that the percentile "is not equal to the percentage correct on the test". Its own example: 38% correct on a school test is a failing grade, but the 38th percentile is within one standard deviation of the average.

The comparison group matters as much as the score. The American Psychological Association explains that a norm-referenced test may rank a child's ability "compared to other children of similar age or grade level" (APA, page updated August 2022). Adult tests are normed by age as well: the WAIS-5 sample was built in age bands from 16 to 90 (Winter, Trudel and Kaufman, 2024). Riverside Insights, which publishes the Woodcock-Johnson IV, notes that the same performance can rank differently depending on who you are compared with.

So a percentile answers one question: out of 100 people like the ones the test was normed on, how many would you match or beat?

IQ score to percentile chart

These values are calculated from the normal curve with an average of 100 and a standard deviation of 15. A 2025 study gives that scale for the WAIS-IV composite scores (Albayrak Rasborg Hartogsohn and colleagues), and a 2025 review gives a standard deviation of 15 for the Wechsler scale (O'Sullivan and colleagues). A summary by the Texas Statewide Leadership for Autism Training gives the same scale for the Stanford-Binet Fifth Edition. Real tests publish their own tables, which can differ slightly at the ends.

• 70: 2.3rd percentile
• 75: 4.8th percentile
• 80: 9.1st percentile
• 85: 15.9th percentile
• 90: 25.2nd percentile
• 95: 36.9th percentile
• 100: 50th percentile
• 105: 63.1st percentile
• 110: 74.8th percentile
• 115: 84.1st percentile
• 120: 90.9th percentile, about 1 in 11 people score this high or higher
• 125: 95.2nd percentile
• 130: 97.7th percentile, about 1 in 44
• 135: 99.0th percentile, about 1 in 102
• 140: 99.6th percentile, about 1 in 261
• 145: 99.9th percentile, about 1 in 741

The numbers bunch in the middle. By the normal curve, going from 100 to 110 moves you about 25 percentile points, while going from 130 to 140 moves you about 2. So near the middle a few points move a percentile a long way, and near the top even a 10-point gap moves it very little.

The labels tests attach to these bands are separate from the percentile. On the WAIS-IV, 110 to 119 is High average, 120 to 129 is Superior and 130 and above is Very superior (Hwang and colleagues, Psychiatry Investigation, 2019; Hemmingsen and colleagues, BMC Psychiatry, 2020). The fifth edition of the Wechsler children's test uses very high and extremely high for the same two top bands (Medise and colleagues, Frontiers in Pediatrics, 2024). Our guide to the average IQ explains the bands in more detail.

Percentile to IQ: what score puts you in the top X%

Many people start from the other end: they know a percentile and want the score. By the normal curve, with an average of 100 and a standard deviation of 15:

• 75th percentile (top 25%): 110.1
• 90th percentile (top 10%): 119.2
• 93rd percentile (top 7%): 122.1
• 94th percentile (top 6%): 123.3
• 95th percentile (top 5%): 124.7
• 98th percentile (top 2%): 130.8
• 99th percentile (top 1%): 134.9

How to calculate your IQ percentile

You need two steps and the normal curve.

1. Turn your score into a z-score. A z-score says how many standard deviations you sit from the average. Online Statistics Education (David M. Lane, Rice University) gives the formula: z = (X minus the mean) divided by the standard deviation. For an IQ score, that is (your score minus 100) divided by 15.
2. Look up the z-score on the normal curve. A z-table or any spreadsheet's normal distribution function returns the share of people below that point. Multiply by 100 for the percentile.

A worked example: take a score of 120. The z-score is (120 minus 100) divided by 15, which is 1.33. The normal curve puts 90.9% of people below that point, so a score of 120 sits at the 90.9th percentile.

Two rules of thumb come from the same curve. The NIST/SEMATECH e-Handbook of Statistical Methods, from the US National Institute of Standards and Technology, gives 68.27% of a normal population within one standard deviation of the average, 95.45% within two and 99.73% within three. On the IQ scale those bands are 85 to 115, 70 to 130 and 55 to 145. The US Supreme Court used the same scale in Hall v. Florida (2014): "The mean IQ test score is 100" and the standard deviation "is approximately 15 points".

Why the same percentile has different numbers on different tests

Not every test uses a standard deviation of 15. The fourth edition of the Stanford-Binet reported its composite with an average of 100 and a standard deviation of 16, according to a 2024 study by Marteleto and Perissinoto in the Sao Paulo Medical Journal. A 2022 paper in Autism Research traces that standard deviation of 16 back to the 1960 Stanford-Binet revision. The Fifth Edition moved to 15 (Texas Statewide Leadership for Autism Training, SB5 summary).

American Mensa's list of qualifying scores shows what this does to the numbers (us.mensa.org, read 16 October 2026). It accepts 130 on the Wechsler adult and children's scales and on the Stanford-Binet Fifth Edition, and 148 on the Cattell. It lists a separate Stanford-Binet entry at 132; by our reading that is an edition before the fifth, which used a standard deviation of 16. By the normal curve, two standard deviations above the average is 130 on a scale with a standard deviation of 15, 132 on 16 and 148 on 24. So the three numbers are the same distance from the average on three different rulers.

Sources do not fully agree on the Cattell scale. A 2025 review lists the Culture Fair Intelligence Test with a standard deviation of 16 (O'Sullivan and colleagues, Orphanet Journal of Rare Diseases), while by the normal curve Mensa's cutoff of 148 only fits a scale of 24. Before you compare two scores, check which scale each report names.

Your score is a range, so your percentile is too

No test measures anyone to the exact point. In Hall v. Florida (2014), the Supreme Court wrote that experts "have agreed, for years now, that IQ test scores should be read not as a single fixed number but as a range". The opinion quotes the APA's brief: "1 SEM equates to a confidence of 68% that the measured score falls within a given score range, while 2 SEM provides a 95% confidence level". Its example is a score of 71, read as 66 to 76 with 95% confidence.

SEM stands for standard error of measurement. Assessment Systems Corporation, a psychometrics software firm, gives the classical formula: SEM = SD × sqrt(1 − r), where r is the test's reliability (page updated August 2026). Winter, Trudel and Kaufman (2024) report that the WAIS-5 FSIQ reliability "averaged 0.97 across the age range". By that formula, a reliability of .97 on a scale with a standard deviation of 15 gives an SEM of about 2.6 points, so a 95% range of about plus or minus 5 points. Less reliable tests have wider ranges.

Here is what a range does to a percentile. By the normal curve, a score of 120 read as 115 to 125 spans the 84.1st to the 95.2nd percentile: about 11 percentile points. The court also listed reasons a score moves between sittings, including "the test-taker's health; practice from earlier tests" and "simple lucky guessing". Our article on whether IQ tests are accurate covers how much scores change on a retake.

Why percentiles at the very top and bottom are less precise

The far ends of the chart rest on few people. The WAIS-5 was normed on 2,020 people aged 16 to 90, with 180 per age band for ages 16 to 69 (Winter, Trudel and Kaufman, Journal of Intelligence, 2024). By the normal curve, about 1 in 741 people score 145 or higher, so a band of 180 would be expected to hold fewer than one of them.

Tests also have floors and ceilings. A 2023 study states that the Wechsler children's FSIQ "ranges from 40 to 160" (Rosolen and colleagues, Toxics). Sansone and colleagues (Journal of Neurodevelopmental Disorders, 2014) write that "Most intelligence tests do not measure IQ below 40". The Stanford-Binet Fifth Edition has an extended scale that can go beyond 160 and below 40 (SB5 summary). Pearson's WAIS-IV FAQ adds that reliability is lower in gifted and intellectually disabled samples, because "the restriction in the range of scores obtained by these groups frequently results in lower reliabilities".

So a report that gives a top-end percentile to two decimal places is giving more precision than the test can support. Past about 145, the useful statement is that the score is very high, with its range.

Old norms make percentiles read high

A percentile is only as current as the norm group behind it. Writing in the Journal of Intelligence in 2024, Winter, Trudel and Kaufman say that clinicians became aware that "IQ test norms become out of date at the rate of three IQ points per decade", a drift known as the Flynn effect. They add that "Older norms tend to produce spuriously high IQs".

The size of the drift depends on the study. Trahan and colleagues' 2014 meta-analysis in Psychological Bulletin, across 285 studies, found a mean of 2.31 points per decade, and 2.93 for modern Stanford-Binet and Wechsler tests. Winter and colleagues (2024) found only 1.2 points per decade in two relatively small counterbalanced studies from the WAIS-5 manual: 186 people tested on the WAIS-IV and the WAIS-5, and 98 people aged 16 tested on the WISC-V and the WAIS-5. Pearson's WAIS-IV FAQ says "the norms for the newer test are slightly harder due to the Flynn effect". If your score comes from an old edition, its percentile probably reads a little high.

What an online test's percentile can tell you

An online test's percentile compares you with that test's own norm group, whoever that is. The Leaders Project explains that the standardization sample determines "what is considered normal or typically developing performance for that test", and that "Most commercially available tests use a standardization sample that reflects the most recent U.S. census information". The WAIS-5 sample was matched to 2022 US Census proportions by age, sex, ethnicity, education and region (Winter and colleagues, 2024).

For examinees who are English language learners, Pearson's WAIS-5 page says the summary report "should include a caveat that the WAIS-5 was normed on a sample of primary English speakers". American Mensa states: "We do not accept unsupervised testing as proof of eligibility, specifically unsupervised testing administered electronically, or via internet-based tests." Our inference from these sources: an online percentile is a useful estimate, and how much it tells you depends on how its comparison group was built.

A clinical assessment is a different thing. The APA explains that a psychological assessment can combine tests with interviews, records, observation and other information, with a psychologist deciding what information to use. If you need a score for a school, a clinic or a legal decision, ask a psychologist. Our guides to the WAIS and the Stanford-Binet describe what those tests involve.

How to read the percentile on your own report

• Find the scale. Check the test name and its standard deviation before you convert anything.
• Find the comparison group. Look for the age group and the year the norms were collected.
• Read the range as well as the point. Convert both ends of the range to percentiles.
• Compare like with like. Two scores from different tests or scales can mean the same position.

WhatIQ's test has 24 timed picture puzzles and takes about 15 minutes (whatiq.com, 16 October 2026). Its report gives a score on the standard IQ scale with a percentile and a 90% range, and the site says plainly that its result is an estimate and is not a clinical assessment.

Frequently asked questions

What percentile is a 120 IQ, and is it high?

By the normal curve with an average of 100 and a standard deviation of 15, a score of 120 is at the 90.9th percentile, and about 1 in 11 people score that high or higher. The WAIS-IV labels 120 to 129 as Superior. Your true score sits in a range around 120, so the percentile is a range too.

What IQ is the top 1 percent?

By the normal curve, the 99th percentile sits at 134.9, so about 135, on a scale with an average of 100 and a standard deviation of 15. On a test that uses a standard deviation of 16, the same position takes a higher number. Scores this high are less precise, because few people in any norm group reach them.

What IQ is in the top 7 percent?

The top 7% starts at 122.1 on the standard scale, by the normal curve with an average of 100 and a standard deviation of 15. The top 6% starts at 123.3 and the top 5% at 124.7. Because each score carries a range of several points, two people a point or two apart may not differ at all.

How rare is a 126 IQ?

By the normal curve with an average of 100 and a standard deviation of 15, a score of 126 sits at about the 95.8th percentile. About 1 in 24 people score that high or higher. Treat it as an estimate: a score read as a range of a few points either side covers a wider span of percentiles.

What percentile do you need for Mensa?

American Mensa asks for a score in the top 2%, at or above the 98th percentile on an approved, supervised test. Its list accepts 130 on the Wechsler scales and the Stanford-Binet Fifth Edition. By the normal curve, 130 is the 97.7th percentile, so the strict 98th percentile point is 130.8. Mensa does not accept internet-based tests.

Is an IQ percentile the same as the percentage of questions right?

No. A percentile compares your score with other people in the norm group, while percent correct counts your answers. The Leaders Project at Teachers College, Columbia University gives the example that 38% correct would fail a school test, while the 38th percentile is within one standard deviation of the average, close to the middle of the group.