Celebrities With the Highest IQ: Where Each Number Came From

0 Quick Answer: Four Evidence Classes, Not One Ranking

There is no reliable ranking of celebrities by IQ, because almost none of the numbers came from a test. Across the thirty five famous people examined in detail on this site, four have any cognitive administration on record at all, and only one of those four has a named instrument, a named date, a named venue and a comparison group tested alongside him. Every other figure is an inference, a self report, or a number with no origin anyone can name.

So this page does not rank. It sorts. Every claimed celebrity IQ belongs to one of four evidence classes. Documented administration means a real test was given and a record of it survives. Documented proxy means no IQ test exists, but a school exam, an admissions test or a military qualification battery does, and it can be mapped onto an IQ scale if the assumptions are stated out loud. Published expert estimate means a researcher applied a stated method, usually historiometric, and published the result along with its limits. Unsourced folklore means the number circulates and nobody can say where it started.

Sorting this way changes which cases are interesting. Garry Kasparov, who was actually tested, scored lower than the folklore attached to chess champions. Bill Gates, who left a hard proxy behind, supports a smaller number than the one people repeat. Richard Feynman's reported figure is famously modest. The pattern is consistent: wherever evidence exists, the number comes down.

1 The Four Evidence Classes, Defined Precisely

A class is defined by what must exist for the claim to be checkable, not by how confident the person repeating it sounds. The four tiers below are ordered by how much a reader can verify without taking anyone's word for it. Each tier has an entry requirement, and a claim that fails the requirement drops to the next tier down regardless of who published it.

Class one, documented administration, requires four things: a named instrument, a date, an administering party, and a result that someone other than the subject can point to. Class two, documented proxy, requires an authenticated non-IQ score plus an explicitly stated conversion assumption. Class three, published expert estimate, requires a stated method, a named researcher and published uncertainty. Class four is everything else, which in practice means most of the internet.

Two boundary cases deserve their own note, because they get miscategorised constantly. A self report is not class one. When a public figure states a number about himself, that establishes that the claim was made, not that the administration happened as described. Andrew Tate's self reported 148, Kanye West's self stated 133 alongside other figures he has given, and the figure above 150 attributed to Jordan Peterson without a locatable primary source all sit in class four until documentation appears.

The second boundary case is the wrong instrument. A cognitive test that is not an intelligence test cannot be reported as one. The Montreal Cognitive Assessment, described by its publisher as a screening tool for detecting mild cognitive impairment across memory, visuospatial ability, executive function, attention and orientation, is the only cognitive test Donald Trump is known to have taken. It is a dementia screen. Screening for impairment and estimating ability at the top of the range are different measurement problems entirely.

2 Class One: The Very Short List of People Who Were Tested

Four of the thirty five cases involve a real administration, and the strongest of them is not a scientist or a billionaire but a chess player. In December 1987 the German magazine Der Spiegel spent three days putting the reigning world champion through cognitive tasks in a hotel suite in Baku. Two numbers came out: 135 on a battery Hans Jurgen Eysenck assembled for the occasion, and 123 on Raven's Progressive Matrices. There was a named publication, a named place, a named date, named instruments and a comparison sample. The full Kasparov case is the most valuable single data point in the genre, and the reason is uncomfortable: the archetypal genius of twentieth century chess scored high but not mythically high.

The second case is historical and thinner. Feynman's reported 125 comes from a test given in high school and is described in James Gleick's biography. It has a source trail, which puts it above folklore, but the instrument, the norm table, the age correction and the error band are all missing. The honest label is a reported school result, not a modern adult Full Scale IQ.

The third and fourth cases are administrations whose instruments do not support the interpretation people put on them. Joe Rogan's 127 came from a paid online test he took and shared, whose own published terms disclaim validity and describe access as being for entertainment. Christopher Langan's 195 to 210 came from untimed, unsupervised high range instruments normed on people who chose to attempt them. Something was administered in both cases. Neither administration produces a score comparable to one from a normed battery. The same mismatch, with far better paperwork behind it, sits outside the thirty five cases tabulated below: the armed forces mental qualifying examination Muhammad Ali sat in 1964 was a real, dated, official administration that placed him at roughly the 16th percentile, and the figures of 78 and 83 later attached to that rank name no intelligence test at all.

Notice what the class one list does to the premise of the question. If you rank the audited cases by evidence quality rather than by number, the top of the list is occupied by figures in the 120s and 130s. The 190s and 200s all sit at the bottom, in the class with no instrument behind it. That inversion is the single most useful thing this page has to report, and it survives every attempt to explain it away.

One further caution about class one membership. Being tested does not make a number portable. A score is a statement about performance on one instrument, under one set of conditions, against one reference group. Moving it onto a different scale requires the kind of documentation covered in how IQ scores are normed, and none of these four cases carries it.

3 Class Two: Admissions Tests, School Exams and Military Batteries

A proxy is honest only when the conversion assumption is printed next to the answer. The most common proxy in celebrity cases is the SAT, and the reason it is used is that the relationship between it and general cognitive ability is genuinely strong. Meredith Frey, reviewing fifteen years of evidence in the Journal of Intelligence in 2019, summarises the finding from Frey and Detterman's 2004 study that SAT scores correlate up to about .8 with measures of fluid reasoning and g, performing about as well against traditional intelligence tests as those tests do against each other.

Strong at the group level is not the same as precise for one person. Correlations near .8 leave substantial variance unexplained, and the conversion loses accuracy exactly where celebrity cases live, near the ceiling, where a handful of raw points can move an estimate by a wide margin. That is why the Gates case lands where it does. He scored 1590 out of 1600 on the harder pre-1995 SAT, which is a real, hard, documented cognitive result, and he later co-published a bound on pancake sorting that stood for roughly thirty years. Weighing both while correcting for the imprecision of near ceiling conversion, that page reaches an estimated 140 to 150 rather than the 160 people repeat.

The same method applied to Elon Musk pulls in the opposite direction from the folklore, with conversions from his reported older SAT landing nearer the low 120s than the viral 155. Same technique, same discipline, an answer nobody wanted. That is what distinguishes a method from a flattering guess.

Every entry in that table is an estimate produced by weighing a record, not a measurement. The distinction matters enough that it is worth restating in the plainest terms: a conversion from a proxy produces a band, and the band should be reported with its width. Anyone who converts a proxy into a single number without showing the width has skipped the only step that made the exercise defensible. For the mechanics of moving between scales, see how SAT scores relate to IQ scales and what a standard deviation of 15 actually does to a score.

4 Class Three: Historiometric Estimates and What They License

Two published bodies of work supply almost every historical IQ figure still in circulation, and both stated their limits far more clearly than the sites that quote them. The first is Catharine Cox's volume in Lewis Terman's Genetic Studies of Genius series, The Early Mental Traits of Three Hundred Geniuses, whose full text is on the Internet Archive. Cox took 301 eminent figures born between 1450 and 1850 and estimated childhood intelligence from surviving biographical evidence, then corrected upward for how little of a given childhood happened to be recorded.

That correction is the whole method in miniature, and it is the part the listicles drop. Newton's famous 190 is the corrected figure; the estimate from documented achievement to age seventeen was closer to 170. A number that depends on the size of the surviving archive is a statement about documentation as much as about the mind being documented. Cox knew this and said so, which is why her work belongs in class three rather than class four.

The frame also excludes people it is routinely used to cover. Marie Curie was born in 1867, outside Cox's window entirely, so the 180 to 200 figures attached to her have no historiometric basis at all and come from popular lists. The 180 and 220 attached to Leonardo circulate without any stated method behind them. Einstein's 160 is a retrospective attachment to a man for whom no evidence of any IQ administration exists.

The second body of work is Dean Keith Simonton's historiometric estimation of presidential intelligence, published in Political Psychology in 2006 for forty two chief executives from Washington to George W. Bush, using educational records, the linguistic complexity of presidential writing, documented achievements and expert ratings. Simonton revisited the same data in the Journal of Intelligence in 2018, where he reports that Intellectual Brilliance correlated with rated presidential performance at about .56 while the Cox derived IQ estimates managed roughly .31 to .34, and that only Intellectual Brilliance survived in the multivariate model. How little of that ranking was ever observed is visible in the 2006 table itself, where only eight of the forty two presidents carried an IQ figure before the imputation ran, all eight of them from Cox, and the remaining thirty four numbers are algorithm output rather than anything anyone measured.

Simonton's own finding is the useful one for this page. When he put estimated IQ and rated intellectual brilliance into the same model, the estimated IQ stopped mattering. If the researcher who built the most careful estimates in the field concludes that the estimates are the weaker predictor, a celebrity list built entirely from weaker versions of the same exercise is not measuring anything. See the political IQ claims page for how this gets abused in both directions.

5 Class Four: How an Unsourced Number Is Manufactured

The largest class is the one where a figure has no origin, and a few cases let you watch the manufacturing happen. The cleanest is Snoop Dogg's 147, which began as a graphic posted by an entertainment account on social media and spread because the subject himself reposted it. There is a visible birth event: an image, an account, a share, and then several years of articles treating the result as biography. No psychologist, no testing service, no document.

Repetition then does the rest. The 152 attached to Zuckerberg appeared in coverage around 2010 and propagated without ever acquiring a source. The 170 attached to Sam Altman travels through citation loops in which estimate pages cite each other and the chain never terminates in an instrument. Following the 145 attached to Obama leads to other celebrity lists rather than to a test. The trail always ends the same way, which is to say it does not end at all. The 178 attached to Rowan Atkinson runs the same loop with an extra layer on top, because a July 2025 fact check established that the widely shared photograph of him as a teenage engineering student at Newcastle had been generated by AI, so a fabricated image now travels alongside the unsourced number.

Three tells identify a manufactured number quickly. First, contradictory precision: the figures 102 and 130 both circulate confidently for Cristiano Ronaldo, and both 160 and 136 circulate for Taylor Swift. Two exact numbers that disagree by thirty points cannot both be measurements. Second, subject denial: Dolph Lundgren has said he does not know his own score while the 160 has followed him for well over a decade. Third, numerical impossibility: several figures published for Vitalik Buterin are not attainable on any standard scale.

The most instructive case has no folklore number at all. Jimmy Donaldson has never published a score and no rumour figure settled on him, so automated estimator pages generate one from his biography and present the output as a finding. That is the mechanism laid bare. The estimator does not measure; it reads fame and returns a number scaled to it, which is why the outputs disagree with each other.

Two more cases sit in class four for reasons worth naming. The 160 attached to Hawking functions as cultural shorthand for a famous physicist rather than as a score report. Neil deGrasse Tyson has publicly refused the whole exercise for years, noting in 2015 that research continues to demonstrate that your IQ is an excellent predictor of your score on an IQ test, which is a joke with a real thesis inside it. Numbers still get assigned to him by other people.

6 Every Case, Sorted by Evidence Class

Here is the full audit in one place, ordered by evidence class rather than by the size of the claim. Each row links to the page that examines that case in full. Read down the right hand column rather than the middle one: the middle column is what circulates, and the right hand column is what a reader can actually check.

Twenty five of the thirty five rows sit in class four. That ratio, not any individual entry, is the finding. For the separate question of who holds records and where measurement itself runs out, see the highest IQ ever guide, which owns that ground and is not repeated here.

7 How to Read a Score Band Before Browsing One

A band is only meaningful once you know how many people are in it. Searches for famous people with 130, 140 or 160 IQ assume those bands are populated at roughly similar rates, and they are not. On a scale with mean 100 and standard deviation 15, each step upward costs vastly more than the last, and by the time you reach the numbers celebrity lists prefer, the population has almost emptied.

The rarity figures below come from the site's own reference tables rather than from any claim about a specific person. They are the arithmetic of a normal distribution with the stated parameters, and they are what makes the next three sections readable. Anyone quoting a 160 for a public figure is implicitly claiming membership of a group with roughly one member in thirty thousand.

Two properties of that table drive everything that follows. The first is acceleration: the distance from 130 to 145 costs a factor of about seventeen in rarity, while the distance from 145 to 160 costs a further factor of about forty three. The second is that measurement precision moves in the opposite direction. As the population thins, the number of people available to norm the band thins with it, which is why validated instruments stop reporting long before the numbers in celebrity headlines.

Use the percentile chart to convert any score into its exact rank, and the rarity calculator to see the one in X figure and how many people worldwide sit at or above it. For the labels themselves and where the gifted thresholds fall, the gifted range guide and the score interpretation guide cover the ground. For why rank and raw performance are different objects, see score versus percentile.

8 Famous People With a 130 IQ: The Band Estimates Actually Reach

The 130 band is where careful estimates for famous people cluster, and almost nobody searches for it. At 130 a person sits at the 97.7th percentile, roughly one person in every forty four. That is genuinely uncommon and entirely compatible with an ordinary, visible life. It is also, repeatedly, where the audited pages land once folklore is stripped out.

Working through the cases: Obama's estimate is 125 to 135 from an academic and legal record, well short of the circulating 145. Swift's estimate is 120 to 130 rather than the 160 in circulation. Trump's estimate is 120 to 130 against a claimed 156. West's estimate is 115 to 130 against his own reported 133. Tate's estimate is 110 to 125 against a self reported 148. Five different careers, five different kinds of evidence, one narrow band.

That convergence is not a coincidence and it is not modesty. It is what happens when an estimate is anchored to things that can be checked, because the checkable evidence for most public figures consists of educational attainment, professional selection and verbal output, all of which are informative and none of which discriminates well above about two standard deviations. Selection into Wharton, Columbia or a record deal is a filter that many people at 125 pass and few people at 100 do. It is not a filter that distinguishes 135 from 165.

Every figure in the paragraph above is an editorial estimate produced by the linked page from a documented record, not a measurement, and each one is published with an uncertainty band wider than the estimate itself. That is the correct way to report an inference. A single number quoted without its width is making a promise the evidence cannot keep.

For what the 130 threshold means institutionally, note that Mensa International defines admission as scoring within the upper two percent of the general population on an approved and supervised test, which is a percentile requirement rather than a bare number. That distinction is covered in the high IQ society requirements guide, and the band itself in the IQ 130 page.

9 Famous People With a 140 IQ: Where Proxy Evidence Runs Out

The 140 band is the highest one that documented proxy evidence can honestly reach, and even there it reaches as a range. A score of 140 sits at the 99.6th percentile, roughly one in two hundred sixty one. Three of the four proxy cases on this page produce estimates that overlap it: Gates at an estimated 140 to 150, Bezos at an estimated 135 to 148, and Zuckerberg at an estimated 135 to 145. Readers arrive at this band with a label already attached to it, and the label has a single documented origin: the 140 line comes from one suggested classification table on page 79 of Terman's 1916 manual, whose top row reads near genius or genius, and no test in professional use carries a band with that label today.

Notice what those three have in common that the folklore cases do not. Each left behind a hard artefact: a near ceiling score on a harder version of a standardised admissions test, a National Merit selection with a summa cum laude degree in electrical engineering and computer science, a classics diploma from a selective preparatory school with prizes in unrelated subjects. Artefacts constrain estimates. Fame does not.

Two other cases sit near this band with weaker support. Altman's estimate is 125 to 140 from selective education and complex leadership, against a circulating 170. The Jobs estimate is 125 to 140, derived partly from a childhood achievement result that inflates at the extremes, against a circulating 160. In both cases the gap between the estimate and the folklore is roughly thirty points, which is about two standard deviations, which is the distance between the 91st percentile and the 99.87th.

The 140 band also marks a practical boundary for interpretation. Above roughly this point, differences in a general composite stop predicting much about visible outcomes, because the other inputs to a career (opportunity, timing, persistence, capital, health) dominate. That is not a claim that ability stops mattering. It is a claim about how much a single number explains, and it is developed properly in what IQ predicts about job performance and the IQ and success guide. See also the IQ 145 page for the band immediately above.

10 Famous People With a 160 IQ: The Band Where Every Claim Fails

Fifteen of the thirty five audited cases carry a claim at or above 160, and not one of them survives contact with the evidence. A score of 160 sits at the 99.997th percentile, roughly one in thirty one thousand five hundred sixty. If those fifteen claims were measurements, this small collection of famous people would contain a concentration of one in thirty thousand ability that no plausible selection mechanism could produce.

The list is familiar: 160 for Einstein, Tesla, Hawking, Lundgren, Jobs, Gates and Swift; 170 for Altman; 180 or above for von Neumann, Curie, Leonardo, Newton, Carlsen, Langan and Tao. Of these, exactly one traces to a published method with stated uncertainty, and that is Newton's corrected Cox estimate, which its own author treated as an inference from an archive.

There is a second problem beyond sourcing, and it is structural. Standard adult batteries stop reporting well below the numbers on that list, because the norm samples thin out and there is nothing left to calibrate against. A figure of 230 is not a measurement from an instrument with items at that level; it is a model extended into a region where the model was never validated. That is why the Tao case is unresolvable in principle rather than merely undocumented, and why extreme claims cluster around instruments that are neither supervised nor independently reviewed.

The honest conclusion for this band is narrow and it holds across all thirteen. There is no verified celebrity IQ at or above 160. Some of these people may well have scored there had they been tested with an instrument capable of resolving it; nothing here rules that out, and for several of them the achievement record is extraordinary by any standard. What the evidence rules out is quoting the number as a fact. Where measurement itself runs out is the subject of the highest IQ ever guide and the high range test explainer; the band is described in the IQ 160 page.

11 Athletes, Chess Players and Creators: Expertise Is Not a Score

The most common category error in celebrity IQ writing is converting extraordinary domain performance into a general score. The clearest evidence against that conversion comes from chess, the domain people assume is closest to raw reasoning. Burgoyne, Sala, Gobet, Macnamara, Campitelli and Hambrick published a comprehensive meta-analysis in Intelligence in 2016 reporting that chess skill correlated with fluid reasoning at .24, with comprehension knowledge at .22, with short term memory at .25 and with processing speed at .24, an average of about .24, and that the relationship was weaker in adults (.11) than in youth samples (.32).

An average correlation of .24 explains under six percent of variance. Applied to Carlsen, whose rating record is a genuinely well validated measurement of playing strength, the implication is uncomfortable for the 190 attached to him: even a verified general score would explain very little about why he is who he is. Applied to Kasparov, it explains why an actual administration produced 135 rather than the number people expect from a world champion.

The same logic covers the athletes. LeBron James's recall of games and reading of plays is real and elite, and it is trained expertise built over decades rather than a general composite. Messi's case sits next to an actual research literature on football perception, using eye tracking, head movement sensors and occlusion experiments, which measures visual search and anticipation rather than a Full Scale IQ. Ronaldo's estimate is deliberately cautious for the same reason: elite sport reveals exceptional expertise under conditions no battery reproduces.

Creators fit the pattern too. MrBeast's method looks like cognitive work from the outside because it is systematic hypothesis testing at enormous scale, and that is exactly why an estimator page can generate an impressive number from it. Snoop Dogg's documented business record is real evidence of judgment and does not produce a score either.

None of this diminishes the achievements. It relocates them. Broad cognitive ability, in the structure described by the CHC model, is one input among many to elite performance, and the domain specific knowledge built over ten thousand hours is a different construct from the general factor. Treating a career as a score confuses the two, in both directions.

12 Auditing Any Celebrity IQ Claim in Ten Minutes

The audit that produced this page is short enough to run yourself, and it terminates quickly on almost every claim. Six questions, asked in order. Stop as soon as one fails, because a claim that fails an early question cannot pass a later one.

• What instrument? A test has a name and an edition. If nobody can name one, the claim is class four and the audit is over. This eliminates most figures in circulation before the second question.
• What date and what age? Norms are age referenced. A childhood result and an adult result are not interchangeable, which is what makes Feynman's school figure a school figure rather than an adult composite.
• Who administered it? Supervised administration by a trained examiner is what separates a score from a screenshot. This is the question that reclassifies Rogan's result and Langan's.
• Against which reference group? A number without a norm sample means nothing. Instruments normed on people who chose to enter are measuring a self selected group, not the population.
• Where is the interval? Every real score carries measurement error. A claim quoted to the exact point, with no band, has already discarded the information that would let you check it.
• Who is the original source? Follow the citation to its end. If the chain runs through estimate pages citing each other, as it does for Altman and Obama, there is no bottom to reach.

Run those six against any name and the outcome is predictable. Question one eliminates the majority. Question three eliminates most survivors. Question five eliminates nearly everything that is left, because celebrity numbers are always quoted as exact points and real scores never are. The four cases that reach question six on this page do so with major caveats attached. A case from outside this page's roster shows how far a claim can get before it stops: the WAIS-R Full Scale of 136 printed in a 1998 federal competency evaluation names its instrument, its 1 May 1996 administration date and a norm sample of 1,880 Americans tested between 1976 and 1980, then fails question five because the document carries no confidence interval, while the 167 attached to the same person fails question one, the evaluator having recorded that the childhood testing was never made available for review.

The same six questions work on a test you are considering taking, which is the more useful application. An instrument that will not tell you its norm sample, its reliability or its confidence interval is asking for the same credulity a celebrity list asks for. The reliability and validity guide covers what those figures mean, the accuracy guide covers what an instrument must publish, and the online accuracy page covers the specific failure modes of unproctored testing.

13 What a Reportable Score Requires, and Where ACIS Stands

The gap between a celebrity claim and a reportable score is not a matter of degree, it is a matter of documentation. A number becomes interpretable when four things are published together: the structure being measured, the reliability of the composite, the error attached to a single administration, and the reference frame the score is compared against. Every one of those is missing from every claim in class four, and most of them are missing from class one too.

For transparency about what this site does and does not publish: ACIS reports 20 subtests across six primary cognitive domains, with scaled scores at mean 10 and standard deviation 3 and composites at mean 100 and standard deviation 15. The Full Scale IQ composite omega is .9886 and its g loading is .958, with a standard error of measurement of about 1.60 IQ points. The higher order g confirmatory model fits at CFI .9761, TLI .9726, RMSEA .0406, SRMR .0217, with a chi-square of 916.703 on 166 degrees of freedom. The technical analysis set behind those tables is N = 2,750 complete records, and the adult reference frame covers 3,243 English speaking records aged 16 to 90.

Those figures come with limits that have to travel with them. ACIS is unsupervised. It is not a clinical or diagnostic instrument, and it is not appropriate for diagnosis, hiring decisions, accommodations or high IQ society admission. Its normative sample is self selected rather than census based, which means the reference frame is a modelled adult frame rather than a representative one, and every interpretation should be read against that. Publishing the limitation alongside the number is the difference between documentation and marketing.

Applying the same six questions from the previous section to this site is the point of publishing them. The instrument is named and described in the technical manual, the structure in the cognitive domains guide, the norming procedure in the norms explainer, and the composite itself in the Full Scale IQ page. Researchers and practitioners administering to participants can use the ACIS for research workspace, or read the administration overview at the Professional section on the home page.

The closing standard is not this site's invention. The Standards for Educational and Psychological Testing (2014), published jointly by the American Educational Research Association, the American Psychological Association and the National Council on Measurement in Education, require that a score be reported with the evidence supporting its intended interpretation, with its measurement error, and with the population it was normed against. APA testing standards make the same demand of anyone who reports a result. Judged against that requirement, the celebrity IQ genre does not produce weak scores. It produces claims that are not scores at all, and reading them as anything else is the error this page exists to correct.

14 Frequently Asked Questions