Take an online IQ test and measure your cognitive skills, logic, and problem-solving abilities. Get your score, percentile and insights instantly.
Excellent 4.8/5
1 million+ test takers
IIF certified
Most people have a rough sense of what they are good at and a much vaguer sense of how that compares with anyone else. An IQ test replaces the vague part with something specific: a normed, comparable estimate of how you reason against people your own age, reported with the honesty a number like that deserves.
Our assessment covers logical and abstract reasoning, verbal reasoning, numerical reasoning, spatial ability and working memory under time pressure. You get a composite score, a percentile, a confidence range and a breakdown by domain — because the shape of a result is usually more useful than its total.
See where you sit on the distribution, not just what you scored — the comparison, not the figure, is the better reason to take a test.
Find the domains you are strongest and weakest in, and what a wide gap between two of them means.
Get an IIF certificate recording the scale, percentile and date — three of the things any score report should state.
Interrogate the result afterwards with our IQ tools.
An IQ test places your reasoning against the performance of other people your age. It reports a position on a distribution: a useful estimate, with known error, of a narrow band of cognitive skill.
In 1904 the French Ministry of Public Instruction wanted to identify pupils needing extra teaching support. Alfred Binet and Théodore Simon answered in 1905 with a graded scale of about thirty short tasks: naming objects, defining abstract words, repeating digits. Binet denied that it measured a fixed quantity.
The arithmetic came later. In 1912 William Stern proposed dividing mental age by chronological age — the ratio behind the phrase intelligence quotient — and Lewis Terman renormed the scale as the Stanford–Binet in 1916. Our history of IQ testing traces how quickly it was pressed into service for schooling, immigration and army selection.
The ratio formula breaks on adults: mental age stops climbing in step with chronological age in adolescence, so a 40-year-old and a 20-year-old of identical ability produce different quotients. David Wechsler abandoned it — his Wechsler–Bellevue scale of 1939, ancestor of today's WAIS, replaced the quotient with a deviation score, which asks how far you sit from the average of your own age group and rescales that answer so the mean is 100. The 100 is a definition, not a discovery: see the bell curve and what an average score means.
Your score is a rank, not an amount, and it carries measurement error: it reports where you fell among same-age peers who sat the same items, as our percentile calculator shows. What these tests capture is a cluster of reasoning abilities that psychometricians since Charles Spearman have called g. Creativity, drive and character are not measured here, as we discuss at length. See also how IQ tests work and what your score really means.
Classification labels are conventions, not diagnoses. They describe a position on the curve and nothing more — two people in the same band can have very different profiles underneath. The percentages below are the share of the general population expected in each band on a scale with a mean of 100 and a standard deviation of 15; some tests norm to 16 instead, which changes every percentage below.
The top 2% of the population, and the conventional threshold used by Mensa. Roughly 1 person in 44. At this level the composite is best read alongside its confidence interval, because individual items start to run out of ceiling.
About 7% of people, sitting between the 91st and 97th percentile. Typically comfortable with abstraction and multi-step reasoning, and well represented in graduate study and analytically demanding work.
Around 16% of people. Above the middle of the distribution without being unusual. Small differences inside this band are often smaller than the measurement error attached to them.
Half the population sits here, spanning the 25th to the 75th percentile. A score of exactly 100 is the definitional midpoint, so it is the single most common result on any properly normed test.
Roughly 16% of people. Often reflects a particular profile rather than a general limit — a slow processing speed or a weak verbal domain can pull a composite down while other domains sit near the mean.
About 9% of the population falls below 80. An online screening result in this range is not a clinical finding and should never be treated as one. A supervised assessment is the only appropriate next step if the question matters.
Use the percentile calculator to convert any score into its population rank, or the score converter to move a figure between scales. Further reading: ranges of IQ level, the IQ level chart and what counts as a genius score.
These are the figures that circulate, not measurements taken by IQ Metrics. Almost none of these people sat a modern, supervised test; where a number has a traceable origin we say so on the profile, and where it does not, we say that too. A separate piece on where celebrity IQ numbers come from follows several of the best-known figures back towards their source. Treat the wall below as a map of what gets claimed, and follow a card for the evidence behind it.
Every figure above links to a profile that separates the reported claim from the documented record. Start at the Celebrity IQ database, or compare yourself against the pattern with the comparison tool once you have taken the test.
Every profile we hold, with the claimed figure, its earliest traceable source and what the record actually supports.
National average estimates, what the underlying studies actually sampled, and why cross-country comparison is harder than the league tables suggest.
Where historiometric estimates come from, what Catharine Cox's 1926 method can and cannot tell us, and why the famous numbers are so wide apart.
The actors, musicians and athletes whose scores get quoted most, and how the claims travelled from a single listicle to common knowledge.
Convert a score between scales, place it on the curve, project it across age bands, or see how far a repeat result should regress towards the mean.
Longer explainers on how the tests are built, what the evidence supports, and where the popular account of IQ goes wrong.
An IQ score is not a count of correct answers. It is a position: where your performance sits relative to a norming sample of people your own age. Understanding that conversion is the difference between reading your result and misreading it.
Modern IQ scales are deviation scores, fixed by definition: the mean (μ) is 100 and the standard deviation (σ) 15. Scores are fitted to a normal distribution — the familiar bell curve, symmetrical about its centre and thin at both tails.
Because the shape is fixed, the 68–95–99.7 rule applies directly: roughly 68% of people score between 85 and 115, about 95% between 70 and 130, and around 99.7% between 55 and 145 — properties of the model, not tallies of real people.
A raw total carries no meaning on its own: twenty-eight out of forty tells you nothing until you know how others fared — and when those others were measured. The conversion runs in three stages.
Your raw total is placed within those of a reference sample, stratified by age and, where possible, education and region.
That position becomes a z-score: how many standard deviations above or below the sample mean you sit.
The z-score is transformed by z × 15 + 100. A z of +1.00 yields 115; a z of −1.33 yields roughly 80.
This is why a 25-item and a 60-item test can return the same estimate: both report a rank, not a workload, and why how the norming was done matters more than how long you sat. See also our accuracy notes and regression to the mean.
A percentile rank states the share of the population scoring at or below your result. The values below are rounded model estimates, not headcounts. The rarity column expresses each score's tail as a frequency — how uncommon a score of 130 really is.
| IQ score | Approx. percentile | Approx. tail rarity |
|---|---|---|
| 70 | 2nd | About 1 in 44 |
| 85 | 16th | About 1 in 6 |
| 100 | 50th | 1 in 2 |
| 115 | 84th | About 1 in 6 |
| 130 | 98th | About 1 in 44 |
| 145 | 99.9th | About 1 in 740 |
Figures at 145 and beyond are the least trustworthy: few norming samples reach that far out. Explore any value with the percentile calculator, compare bands on the level chart, or take the test.
People search for average IQ by age expecting a table that rises and falls. Because modern tests are age-normed, the average at every age is 100 by construction. What changes across a lifetime is raw cognitive performance, not the normed number.
Each age group is compared only against its own peers. A 25-year-old and a 70-year-old who both score 100 are not performing identically on the raw items; each sits at the midpoint of their own reference group, as our page on IQ and age explains.
Before roughly age 10 a child's score is comparatively unstable: early assessments correlate only modestly with adult results, and extreme early scores drift back towards the average on retesting — see regression to the mean. Treat a young child's figure as provisional; our kids IQ test is written with that caution in mind.
Raw processing speed and working memory are usually reported to peak between the late teens and late twenties, while vocabulary, general knowledge and judgement keep climbing for decades after. The “peak age” question therefore has no single answer, and a composite smooths the divergence out of view.
Raymond Cattell and John Horn drew the distinction that makes the age pattern intelligible, and John Carroll's 1993 survey folded it into the three-stratum hierarchy now called the Cattell–Horn–Carroll model.
Reasoning with novel material: matrix problems, unfamiliar sequences, on-the-spot inference. Raw Gf is generally reported to decline gradually from the twenties or thirties onwards.
Accumulated knowledge and verbal skill: vocabulary, factual recall, learned procedure. Gc typically rises into the sixties and then holds.
Because most full-scale tests blend both, the composite hides the divergence, so tracking IQ scores over time is more informative when the subtests are read separately.
Slower processing speed is a normal feature of ageing, not by itself a sign of dementia.
Strictly timed formats penalise that slowing and can understate reasoning ability.
Well-practised knowledge and verbal ability are typically well preserved into later life.
If you are over 60, favour an untimed format and read the result alongside your own earlier scores rather than against a younger cohort. Our IQ test reports a range, not a single figure, for exactly this reason.
Mensa does not publish a single magic number. Its admission rule is a percentile: you must score at or above the 98th percentile of the general population on an approved, supervised test. The number attached to that percentile changes with the scale the test uses.
Founded in Oxford in 1946. The rule is a percentile, not a number: the top two per cent.
Approx. 130 on SD 15 · 132 on SD 16One step further out on the tail: half as many clear it as clear Mensa's.
Approx. 135 on SD 15One person in a thousand. Here the tests start to run out of ceiling.
Approx. 146 on SD 15Mensa was founded in Oxford in 1946 by Roland Berrill and Lancelot Ware on a deliberately narrow idea: a society whose only entry criterion was measured cognitive performance, with no bar based on profession, politics or background. That criterion still governs admission, which is why Mensa requirements are stated as a percentile rather than a score.
The top 2% means roughly one person in fifty. On a modern SD-15 scale — the convention used by the Wechsler scales and by this test — the 98th percentile sits at approximately 130; on the SD-16 scale historically used by the Stanford-Binet, the same cut-off is approximately 132. One level of performance, two numbers: the difference is arithmetic, not ability.
Our score converter translates between the two scales, the bell curve explainer shows why the tail thins so sharply above 130, and the level chart sets the bands out side by side. Read your own result as a rank rather than a trophy: the percentile calculator estimates your population rank, and our accuracy notes explain why it should carry a margin of error.
Intertel was founded in 1966 and the Triple Nine Society in 1978, the latter named for its threshold of one person in a thousand. Each keeps its own list of accepted instruments, so check the rules of the chapter you are applying to.
| Society | Percentile | Approx. SD-15 score |
|---|---|---|
| Mensa | 98th | ~130 |
| Intertel | 99th | ~135 |
| Triple Nine Society | 99.9th | ~146 |
No. No online result, including this one, is accepted as qualifying evidence. Admission comes through Mensa’s own supervised test, or through approved prior evidence from a supervised administration by a qualified examiner. Unsupervised conditions cannot rule out interruptions, retakes or assistance, so no online instrument can meet that standard.
Three steps, about twenty minutes, and a result you can actually reason with. No sign-up is needed to begin, and nothing is scored until you finish.
Items are drawn across logical, verbal, numerical and spatial reasoning, ordered so difficulty rises as you go. Every test-taker sees the same standardised items under the same timing, which is what makes the comparison meaningful.
Correct answers are weighted by item difficulty, ranked against the norming sample for your age band, and converted into a deviation score on the standard scale — mean 100, standard deviation 15.
The result reports your composite, your percentile, the confidence interval around it and a breakdown by cognitive domain, together with an IIF certificate recording the scale and the date.
A well-constructed online test is a good indicator and a poor substitute. It can measure the same reasoning abilities a clinician would probe, but it cannot verify who is at the keyboard, control your surroundings, or watch how you reached an answer.
That distinction explains most of the disagreement about online testing. The instrument and the setting are different things, and only one is under our control: we build the items, the timing and the norms; you supply the room, the device and the state of mind. More detail sits in how accurate IQ tests are.
Reliability means consistency: would this assessment give a similar result under similar conditions? Validity asks something else: does it measure the construct it claims to? A test can be highly reliable and still measure the wrong thing, which is why what makes an IQ test reliable is never a fact about the questions alone.
Internal consistency asks whether items within one test agree; coefficients for well-normed batteries sit near 0.85–0.95.
Test-retest reliability asks whether the same person scores similarly weeks later.
Construct validity is assessed against established batteries — though whether those capture intelligence at all is a further question.
Every test-taker meets identical items, time limits and scoring rules, which removes examiner variance in rapport, pacing and prompting. Non-verbal formats, in the tradition of Raven's matrices and Cattell's Culture Fair scales, reduce the advantage conferred by language fluency without removing it. Self-administration adds noise of its own, though: interruptions, an unfamiliar device or a noisy room all depress performance in ways a supervised setting would catch, and our approach is set out in why us.
Every score carries a confidence interval. A result of 118 is better read as roughly 113–123: the point estimate is convenient, the band is truthful, so quote both together.
Two further effects distort repeat testing: practice effects can lift a second attempt by a few points with no change in ability, and an unusually high or low first result tends to drift back towards your own average, an artefact explained in regression to the mean.
A supervised assessment adds what an online test structurally cannot: identity verification, controlled conditions, observation of effort, and interpretation set against your history. It also carries diagnostic and legal standing a self-administered result does not.
Anyone can print a figure. What makes a result worth having is knowing what was measured, against whom, and how certain it is.
The last of those is the one a single number cannot carry: every composite sits inside a range, and how wide that range should be is part of the result rather than a footnote to it.
Built by specialists in psychology, psychometrics and educational assessment, drawing on the traditions behind Raven's matrices, the ICAR open item bank and the CHC model of cognitive ability. Every item is normed before it counts.
More than a million completed assessments across more than 150 countries. A large, live sample is what keeps the norms current, and current norms are what stop a score drifting out of date.
You receive a domain-by-domain profile, not a lone composite. Two people with an identical total can have opposite shapes, and the shape is usually the part you can act on.
Precision you cannot support is not a service, it is a decoration. Every score we publish arrives with the confidence range that belongs to it.
The IQ Metrics assessment standardEvery reputable cognitive assessment rests on published research rather than private opinion. Three traditions matter most: Raven’s matrices, the open ICAR item bank, and the Cattell-Horn-Carroll model. Our assessment draws on all three.
John C. Raven introduced the Progressive Matrices in 1936: a grid of abstract shapes with one cell missing, and candidate completions from which you select the one that fits the rule. No prose instructions, no vocabulary, no arithmetic. That austerity is not an aesthetic preference — carrying no verbal or curricular content, matrix items reduce, without eliminating, the cultural and educational loading that burdens vocabulary questions. Factor-analytic studies consistently report that they load strongly on g, the general factor Charles Spearman described in 1904, which is why the format anchors so many culture-fair assessments.
ICAR is a public-domain item bank for cognitive ability, described by David Condon and William Revelle in a 2014 paper in the journal Intelligence. Its governing idea is transparency: where most commercial instruments keep their items proprietary, ICAR publishes them openly, so anyone can check the psychometrics independently. The trade-off is that what is public can be practised — a cost worth paying, and one that shapes who we are as a publisher.
Matrix reasoning: Raven-style abstract completion items.
Verbal reasoning: inference and relational questions in language.
Letter and number series: rule induction over ordered sequences.
Three-dimensional rotation: mental manipulation of solid forms — with consequences for retesting set out in our accuracy notes.
CHC theory synthesises two lines of work. Raymond Cattell and John Horn separated fluid reasoning (Gf) from crystallised knowledge (Gc) between the 1940s and the 1960s; John Carroll then re-analysed more than 400 datasets in his 1993 survey Human Cognitive Abilities and proposed a three-stratum structure. The merged framework now shapes how batteries such as the Woodcock-Johnson are built, as how IQ tests work explains.
| Stratum | What it contains |
|---|---|
| III (general) | A single general factor, g. |
| II (broad) | Gf, Gc, visual processing (Gv), short-term memory (Gsm), processing speed (Gs) and others. |
| I (narrow) | Dozens of specific abilities, each measured by particular task types. |
The number of broad abilities at stratum II is still debated, but the shape is settled: specific skills nest inside broader domains, which correlate with a general factor. That nesting is the argument for reading a domain profile alongside the single figure, which is what our classical battery returns.
Cognitive test scores do carry information about how people fare at work and in education. That information is real, but it is statistical and partial — and what an aptitude score cannot forecast matters more than the headline correlation.
General mental ability, or GMA, is among the most heavily studied predictors in personnel selection: quick, standardised and applicable across many roles. The evidence is strongest for training performance — how rapidly someone acquires new job knowledge — and it strengthens as job complexity rises, so GMA carries more weight for a research scientist than for routine work. Our page on the importance of IQ tests in education and employment sets out where that leaves employers.
Associations between test scores and later educational attainment are consistently reported, and are typically stronger than the equivalent associations with income. Both are real, and both are modest for any individual person.
Childhood scores are associated with years of schooling and examination results. School quality, family resources and continuity of attendance account for much of that association.
Correlations between adult IQ and income vary by study and country, and are weaker than the schooling association. Occupation, credentials, geography and inherited advantage all intervene.
A correlation across thousands of people tells you very little about one person. Two people with identical scores routinely end up in very different circumstances.
The variance cognitive scores fail to explain is the clear majority of it.
Restriction of range weakens prediction inside already-selected populations, such as graduate intakes.
Conscientiousness, health, opportunity and timing all contribute independently — and personality measures something ability tests do not.
Teams benefit from varied thinking styles, as explored in cognitive diversity at work.
Prediction here is probabilistic, never destiny. For the balanced case see the pros and cons of IQ testing and what your score really means, then take the test and treat the number as information.
Cognitive ability is one of the more consistently studied predictors of training success, particularly in complex roles. It is not destiny — but understanding your own profile helps you choose work that plays to how you actually think.
A pre-employment aptitude test is short, strictly timed and deliberately mixed: verbal, numerical and logical items in one sitting. The two best-known are the Criteria Cognitive Aptitude Test (CCAT), with 50 items in 15 minutes, and the Wonderlic Cognitive Ability Test, with 50 items in 12 minutes. Very few candidates finish either, and that is by design: these are speeded screens measuring how much correct work you produce under pressure, not clinical assessments.
Two people can reach the same composite score by entirely different routes, and the shape of the profile is usually the more useful career signal — which is why what your score really means depends on the breakdown.
Mental rotation and abstract reasoning suit engineering, architecture, surgery and industrial design: roles that reward holding a structure in mind. Our spatial reasoning test isolates it.
High verbal comprehension suits law, editorial work, policy analysis and teaching: roles where meaning must be extracted precisely, then rebuilt. A verbal reasoning test reads cleaner than a composite.
Numerical fluency suits actuarial science, quantitative finance and empirical research. Comfort with proportion matters more than speed. Try the numerical test, then your subtest comparison.
Abstract reasoning is not numeracy, and conflating the two costs people opportunities. Pattern-finding, analogical thinking and verbal fluency are what drive strategy, design, research direction and most genuinely creative work. A lopsided profile is an asset when matched to the right role. Test the components separately across the full assessment hub, beginning with the standard assessment to establish a baseline.
Excellent 4.8 / 5
from over 1 million completed assessments
I had taken two other online tests and got two very different numbers with no explanation. This was the first one that told me the scale it used and how much error to expect. That single detail made the result feel like information rather than a verdict.
My composite was around what I expected, but the spatial and verbal scores were much further apart than I would have guessed. It explained something about the kind of work I find easy that I had never been able to put into words.
Started it on my phone during a commute and finished in about twenty minutes. No account needed before I began, no pressure at the end, and the certificate arrived with the date and scale on it, which is what I actually wanted.
I appreciated being told plainly that an online result is an indicator and not a clinical assessment. I have seen sites imply the opposite. That candour is the reason I trusted the rest of the numbers.
Second result was four points off the first, comfortably inside the range they had given me the first time. Seeing the prediction actually hold did more for my confidence in the test than any marketing copy would have.
The questions we are asked most, answered plainly. If yours is not here, the full FAQ goes further, and contact us if you would rather ask directly.
Taking the testAbout twenty minutes for most people. There are 30 questions, and while individual items are timed, you are not racing a single global clock. Finish it in one sitting if you can: interruptions add noise to a measurement already sensitive to fatigue and attention.
No. There is no signup and no email address required to begin — you can start immediately and work through all 30 questions. At the end you unlock your official score, your printable IIF certificate and the full performance report, and a copy of the result is sent to your email.
No, and cramming does not help — the items test reasoning rather than recall. What genuinely helps is the boring advice: a full night's sleep, a quiet room, a device you are comfortable with, and taking it when you are alert. Our guide on how to prepare for an IQ test covers this in more detail, and whether the underlying ability itself can be changed is a separate question.
Yes. Bear in mind the practice effect: familiarity with the item formats tends to lift a second score modestly even when nothing about your reasoning has changed, and a very high or very low first result will tend to move back towards the average. Our page on regression to the mean explains why, and IQ over time shows what genuine change looks like.
The main assessment is normed for adults. A child's score is only meaningful when compared with same-age peers on age-appropriate items, so use the kids' IQ test instead, which has its own norms and shorter blocks. Where a formal question about a child's development is at stake, a supervised assessment is the right instrument.
Your official IQ score, your percentile ranking against the general population, a printable IQ certificate issued in your name, and a detailed performance report showing your strengths and weaknesses across each cognitive category.
100 is the definitional average, and roughly two-thirds of people score between 85 and 115. Above 115 places you in the top 16%, and 130 or above in the top 2%. “Good” depends entirely on what you want the number for, which is why we report a percentile and a confidence range alongside the composite. See what an average IQ score means.
Not in any straightforward sense. A higher score means you ranked higher on a specific set of reasoning tasks under test conditions — nothing more. It does not predict happiness, judgement, creativity or decency; on judgement in particular, see why clever people still make bad decisions. Beyond about 130, differences between scores are frequently smaller than the measurement error attached to them, so treating 137 as meaningfully better than 133 misreads the instrument.
Your raw correct answers are weighted by item difficulty and ranked against the norming sample for your age band. That rank becomes a z-score — how many standard deviations you sit from the sample mean — which is rescaled to the familiar 100/15 scale. An IQ of 115 sits one standard deviation above the mean, or about the 84th percentile. You can run the conversion yourself in the percentile calculator.
Yes, and it is one of the most useful things a domain breakdown tells you. A composite is an average across several abilities, so one person can reach 120 through strong spatial and abstract reasoning with weaker verbal scores, while another reaches the same 120 from the opposite direction. That is why we report the domains rather than only the sum.
It affects the comparison, not the arithmetic. Because scores are normed within age bands, the average is 100 at every age by construction — a 60-year-old scoring 110 is compared with other 60-year-olds, not with 25-year-olds. What does change with age is raw performance: processing speed and fluid reasoning decline gradually, while accumulated verbal knowledge holds up. See IQ and age.
For adults, cognitive ability is reasonably stable, so a score from a well-normed test remains a fair guide for years. Two caveats: norms are periodically re-standardised, so an old figure may not be directly comparable to a current one; and childhood scores are considerably less stable. More on how long an IQ test stays valid.
A well-constructed online test is a good indicator and a poor substitute for a supervised clinical assessment. Standardised items and scoring remove examiner variance, but self-administration adds noise a clinic controls for: interruptions, device differences, fatigue, and no verification of who is answering. Read the result as a band rather than a point, and see IQ test accuracy.
No. Mensa requires either its own supervised admission test or approved prior evidence from a supervised, proctored administration, and no online result qualifies — including ours. What a well-calibrated online score does tell you is whether sitting the supervised test is likely to be worth your time. The Mensa section above sets out the thresholds.
Your responses and your score are yours. We do not sell personal data, and a result is never published or attached to your name anywhere public unless you choose to share it. You can begin without creating an account, and you can request deletion of your data at any time. The full detail is in our security and privacy policy.
No. This is an educational and self-insight instrument. It is not a diagnosis, it is not a medical or psychological evaluation, and it should not be used for clinical, educational-placement or employment decisions. If a formal assessment is needed, a qualified psychologist administering a supervised battery is the appropriate route.
Thirty questions. About twenty minutes. A score reported with its scale, its percentile and its confidence range — plus an IIF certificate recording all three.
Start the IQ Test →
