What Is the Highest IQ Ever Recorded?

The highest IQ figure ever listed by a record-keeping body is 228, entered for Marilyn vos Savant in Guinness World Records and printed in every edition from 1986 to 1989.

That is the factual answer. The useful answer is that the number does not mean what almost every article about it implies, that Guinness stopped publishing the category for exactly that reason, and that the larger figures circulating for other people are not measurements at all.

The only officially recorded figure

Marilyn vos Savant sat an adult-level Stanford-Binet in September 1956, aged 10. She produced a mental age of 22 years and 11 months. The Stanford-Binet of that era used a ratio IQ, calculated by dividing mental age by chronological age:

22.9 years of mental age ÷ 10 years of actual age × 100 = 229

That reproduces the published 228 almost exactly. And it shows what the number is: not a measure of how far above other adults she scored, but a statement that a ten-year-old performed like a young adult, expressed using a formula that testing abandoned in 1939 when Wechsler introduced the deviation IQ used today.

There is no modern equivalent, because a contemporary test would simply have reported that she was above the ceiling of the childhood norms. Guinness retired the category in 1990 on the grounds that the numbers were too inexact to rank, and it has never returned.

The names you will see, and what the numbers are

Search this question and you will find listicles ranking people by IQ to the nearest point. Here is the same set with the provenance attached, which is the part those lists leave out:

Not one of these is a verified adult score on a modern, properly normed test, because no such score in this range exists for anyone. That is not a criticism of the people listed, several of whom have careers that speak for themselves entirely without a number attached. It is a statement about what IQ tests can and cannot do.

Why extreme scores cannot be trusted

The reason is arithmetic about who was in the room when the test was built, and it is more damning than most people realise.

A test is normed by giving it to a standardisation sample, typically around 2,200 adults, and working out what each raw score corresponds to. Run the normal distribution across that sample and you get:

The distinction between a 140 and a 145 already rests on the performance of about five people. Above 160 there is nobody in the sample at all, so any number reported up there is extrapolation rather than measurement. Responsible tests acknowledge this by reporting “160 or above” instead of inventing a figure.

Scales also diverge sharply at the tails. A score of 160 on a scale with a standard deviation of 15 is the same percentile as about 164 on one using 16, and roughly 196 on the Cattell, which uses 24. Near the average those three barely differ. Out here they are tens of points apart, which means a bare number with no test named is close to meaningless. The mechanics are set out in our guide to an IQ of 140.

The child prodigy arithmetic problem

Nearly every headline IQ figure above 200 has the same origin: a young child, tested with the ratio formula.

The formula rewards precocity enormously, because the denominator is small. A six-year-old performing like a twelve-year-old scores 200. A five-year-old performing like an eleven-year-old scores 220. Neither number says anything about how that person will compare with other adults, which is what a modern IQ score measures.

This is why the formula was abandoned. It also breaks down completely for adults, since chronological age keeps rising after mental ability plateaus, meaning adult scores would decline by arithmetic alone. Wechsler replaced it in 1939 with the deviation IQ, which compares you with age-matched peers on a normal curve and has no equivalent way of generating a 228.

What is the highest IQ actually possible?

There is no theoretical ceiling on human ability. There is a very real ceiling on measurement, and they get confused constantly.

• Up to about 145 to 160, well-built professional tests remain interpretable, with widening error bands as you climb.
• Above that, tests stop discriminating, because they run out of sufficiently hard items and out of people to norm against.
• Around 170 to 190 is where most researchers place the realistic limit of meaningful measurement, and no test in professional use can assign a reliable score above roughly 180.

To see why 200 is not a coherent target, consider what it implies. An IQ of 200 is about 6.7 standard deviations above the mean. On a normal distribution that corresponds to roughly one person in 76 billion, which is nearly ten times the population of the planet. The honest reading is not that such people are vanishingly rare but that the scale has stopped describing anything real that high up, and no instrument could distinguish a 200 from a 180 in any case, because there is no reference group against which to calibrate the difference.

Why the lists keep circulating

Because a ranked list of the smartest people alive is a far better story than an explanation of norming samples, and because the numbers arrived before the caveats did.

Guinness published the category for years, which conferred an authority the figures never had. Once a number is attached to a name in print it propagates, and each republication strips a little more context until what remains is a bare integer that looks like a measurement. The retirement of the category in 1990 was never going to travel as far as the listings had.

The same mechanism explains why Einstein is so often assigned a 160 despite never having taken a test, and why Terman’s 1916 classification row is still quoted as though 140 were an official genius threshold. Numbers outlive the instruments and the caveats that produced them, which is worth remembering whenever you see a confident IQ figure with no test named beside it.

What we can actually say

Stripping out what cannot be supported leaves a smaller but sturdier set of claims:

• Exceptional cognitive ability is real, measurable at the level of “far above the 99.9th percentile”, and genuinely rare.
• It cannot be ranked to the point. Whether someone is at 165 or 185 is not a question current instruments can answer.
• Ability at that level still predicts achievement rather than guaranteeing it. Terman’s own study of high-scoring children produced no Nobel laureates, while two children he screened out later won one.
• Which means the interesting question about anyone on those lists is what they did, not what they scored. In most cases the work is documented and the score is not.

Frequently asked questions

The bottom line

The highest IQ ever officially recorded is 228, and it is a ratio score from a ten-year-old sitting a test in 1956, calculated with a formula that no longer exists in testing. Guinness withdrew the category in 1990 because figures like it cannot be compared. Everything above it is estimate, extrapolation or assertion.

The honest summary is that IQ tests are good at telling you where someone sits among ordinary people and poor at telling you anything precise about the extraordinary. Somewhere past about 160 the instrument stops measuring and starts guessing, and the confident integers you find in ranked lists are downstream of that.