IQ test
A free, research-based IQ test: matrix reasoning, number series, mental rotation and verbal analogies. Items are generated fresh for every attempt, so no answer key exists. You get an estimate with a confidence interval, not a bare number.
Choose a format
Quick
12 items · 8 minutes
Matrices, number series and rotation. No language involved, so results compare across languages. A rougher estimate, with a wider interval.
Start · 12Full
28 items · 25 minutes
All four subtests, including verbal analogies. More precise and gives a breakdown by ability. Best taken somewhere quiet in one sitting.
Start · 28What this test does not do
This is an online screening test, not a clinical assessment. It diagnoses nothing and is not suitable for hiring, school placement or any formal evaluation. Real intelligence testing is done in person by a psychologist using normed instruments such as the Wechsler scales. The number you see is an estimate with error attached, not a verdict and not a description of who you are.
How the test works
The items are not taken from existing sets. Raven’s matrices, the Wechsler scales and CFIT are under copyright and their items cannot be copied, which is exactly what most free IQ tests online do. Here items are built by a generator following published methodology: transformation and logic rules combined the way the Sandia Matrix Generator does, where across a bank of 840 generated matrices item difficulty matched the original scale.
That has a practical consequence: every attempt uses its own set of items. No answer key exists or can exist, so results cannot be looked up, and retaking gains you nothing through memorisation.
Scoring uses the Rasch model: the probability of a correct answer depends on the gap between a person’s ability and an item’s difficulty. This matters because everyone gets a different set: eight correct out of twelve on a hard set and eight on an easy one mean different things, yet the percentage correct is identical. The estimate is the mean of the posterior distribution, so it stays finite even for all-correct or all-wrong responses and always carries a measurement error.
Converting to the familiar scale with a mean of 100 and a standard deviation of 15 currently rests on theoretical item difficulties rather than our own sample. While attempts are few this is a provisional calibration, which is why a range is always shown alongside the score rather than a single number. As anonymous results accumulate, the scale is refined against the real distribution.
Sources
Frequently asked questions
It is a real ability test with scientifically grounded item construction and a psychometric scoring model, but it is not a clinical instrument. An official IQ figure comes from a psychologist working in person with normed scales. An online test gives an estimate with error, and that is why a range is always shown here.
They are not supposed to match. Items are generated afresh for each attempt; only the subtest composition and the difficulty layout are fixed. Results stay comparable because the Rasch model puts different sets on one scale.
You cannot drill the answers, because they do not exist. A small gain on a retake usually reflects familiarity with the format rather than a change in ability. That is true of tests like this generally, not a quirk of this one.
Your answers are never stored and stay in your browser. If you tap "Send anonymously", only the final estimate, the age group and the test version are sent: no answers, no link to you, no account. This is what makes it possible to build a real distribution and refine the scale.
Without a limit it would measure persistence: almost any matrix can be brute-forced given long enough. The limit is generous, roughly twice the usual pace.
It tells you how you handled particular items under time pressure, and nothing about your worth, character or prospects. Tiredness, rushing, an unfamiliar format and your surroundings all affect it. A difference within the range shown is not a difference at all.
Best ideas, in your inbox
Once a week, 3-5 articles actually worth reading. No spam, unsubscribe in one click.