How much IQ does ChatGPT have - and are you smarter than it?

Ever since models like ChatGPT started solving tasks once treated as benchmarks of human intellect, one question keeps flooding the internet: how much IQ does artificial intelligence have, and is it smarter than a human? Plenty of numbers circulate - 150, 160, even 190. Let's see what stands behind them.

Short answer

Artificial intelligence has no official IQ. IQ tests were designed for humans - calibrated on human populations against a human average of 100. Applying that scale to a machine is an analogy, not a measurement. Numbers like "GPT has an IQ of 160" are estimates based on how models perform on reasoning tasks, not the result of a real, standardized assessment.

Where the 150-170 figures come from

In early 2026, analyses and benchmark discussions increasingly place leading models in a range that would correspond to very high human scores - roughly 150-170 "points." Why?

Researchers feed models items from classic ability tests: number sequences, verbal analogies, pattern matrices, spatial rotations. When a model solves them at a level that in a human would match a score of 160, it is tempting to assign it that "IQ." It is a shorthand: we describe a machine's performance using vocabulary built for people, because we have nothing better.

The problem is that these estimates are fragile and inconsistent. The same model can shine on one benchmark and fail a task any child would solve without thinking. In humans, IQ is relatively stable across task types - in AI, it can swing dramatically.

What an IQ test measures in a human

To understand the difference, it helps to recall what an intelligence test actually assesses. A well-built test estimates several related abilities:

• Fluid reasoning - solving novel problems without ready-made knowledge.
• Crystallized knowledge - accumulated vocabulary and facts.
• Working memory - how much information you can hold and manipulate at once.
• Processing speed - how efficiently you analyze stimuli.

In a human, these abilities are intertwined and grow out of a single brain with limited energy. A language model works differently: it has no working memory in the human sense, it does not tire, and its "knowledge" is encoded in billions of parameters learned from text.

What AI does not do the way you do

A model generates the most probable continuation based on patterns in data. That is a powerful tool - but it differs from human thinking in several important ways:

• Understanding versus prediction. A human grasps meaning; a model predicts the next words. Often the output is the same, but the mechanism is not.
• Transfer. You learn a rule from one example and apply it in a completely new situation. A model usually needs an enormous number of examples.
• Awareness and purpose. There is no being here that "wants" to solve the task. No experience, no intention.
• Common sense. Models still stumble over simple questions about the physical world that are obvious to a person.

So who is smarter?

That is a badly posed question - like asking who is faster, a car or a swimmer. It depends on the track. On reasoning benchmarks and raw text-processing speed, the machine outpaces the human. In creativity rooted in lived experience, in understanding another person, in learning from a single experience - you still come out ahead.

Even in humans, IQ is an imperfect measure: it does not capture wisdom, empathy, or creativity in the fullest sense. Stretching it onto machines says more about our fascination with numbers than about intelligence itself.

And how many points do you have?

If you are already wondering about the "IQ" of artificial intelligence, the more interesting question is: where do you land yourself? Unlike a model, your score can be measured on a scale built precisely for humans - with a percentile and interpretation. It takes a few minutes and is free.