Ilya Sutskever's IQ: What Is Actually Known

No published score

Ilya Sutskever has never disclosed an IQ result, and no biography, interview or news report contains one. Any number attached to his name online is unverified. He rarely discusses his personal history in public, and we found no statement from him about intelligence testing.

What the record does contain is unusually precise: dates of degrees, a list of papers with measurable results, and formal honors from scientific bodies. This page uses those, because they are the only evidence that can be checked.

An accelerated start

Sutskever was born in 1986 in Gorky, in the Soviet Union, the city now called Nizhny Novgorod. At 5 he moved with his family to Israel and grew up in Jerusalem. By eighth grade he was taking classes at the Open University of Israel, a distance-learning university with open admission.

At 16 the family moved to Canada. After a brief stint in high school, he was admitted to the University of Toronto as a third-year undergraduate. He earned a bachelor's degree in mathematics in 2005, a master's in computer science in 2007 and a PhD in 2013 under Geoffrey Hinton, with a thesis titled Training Recurrent Neural Networks.

Skipping two years of an undergraduate program at a leading university is a documented marker of advanced ability. It is not an IQ score, and it tells us nothing precise about one. It does tell us that a major university judged a 16-year-old ready for third-year work.

AlexNet, 2012

The result that made Sutskever's name came from a bedroom. As a graduate student in Hinton's lab, he persuaded fellow student Alex Krizhevsky to take on the ImageNet challenge, a contest to classify photographs into 1,000 categories. Krizhevsky trained the network on two Nvidia GTX 580 graphics cards at his parents' house.

In the 2012 ImageNet Large Scale Visual Recognition Challenge, AlexNet posted a top-5 error rate of 15.3 percent, more than 10.8 percentage points better than the runner-up. The paper by Krizhevsky, Sutskever and Hinton, presented at the NIPS conference that December, had been cited more than 198,000 times on Google Scholar by early 2025. It is widely treated as the start of the deep-learning era. Our history of artificial intelligence places it in context.

Sequence to sequence and the Google years

AlexNet was not a one-off. After a two-month postdoctoral stint with Andrew Ng at Stanford, Sutskever joined Google Brain in 2013, when Google acquired DNNresearch, the company he had formed with Hinton and Krizhevsky. There, with Oriol Vinyals and Quoc Le, he wrote “Sequence to Sequence Learning with Neural Networks,” posted in September 2014. The model reached a BLEU score of 34.8 on the WMT'14 English-to-French translation benchmark and set a template that later language systems built on.

He also worked on TensorFlow, Google's machine-learning library, and is among the many co-authors of the AlphaGo paper. Two landmark results in three years, in different problem areas, is the kind of record that separates a lucky strike from sustained ability.

OpenAI and Safe Superintelligence

In 2015 Sutskever co-founded OpenAI and became its chief scientist, the role he held until he left in May 2024. The same year MIT Technology Review named him to its list of 35 Innovators Under 35. In 2022 the Royal Society elected him a Fellow, one of the oldest and most selective honors in science.

On June 19, 2024, he co-founded Safe Superintelligence Inc. with Daniel Gross and Daniel Levy, a company with offices in Palo Alto and Tel Aviv that describes its goal as building safe superintelligent AI. He became its chief executive in July 2025.

Sutskever beside Hinton, Turing and Buterin

His doctoral adviser offers the nearest comparison. Geoffrey Hinton shared the 2018 Turing Award with Yoshua Bengio and Yann LeCun, and the 2024 Nobel Prize in Physics with John Hopfield “for foundational discoveries and inventions that enable machine learning with artificial neural networks.” Neither Hinton nor Sutskever has published an IQ score, so a ranking is impossible. What can be said is that the student co-authored the result that vindicated the teacher's decades-long bet.

Historical comparisons are looser still. Alan Turing and John von Neumann built the theory and machines that made computing possible; their IQ figures are estimates, not scores. Vitalik Buterin is the closest contemporary parallel: born in Russia eight years later, raised in Canada, and at the center of a new field before 25. Browse more computer scientists for other points of reference.

What a score would not show

The most important decision in Sutskever's early career was a bet on large neural networks trained on large datasets, made while much of the field doubted the approach. An IQ test cannot measure that kind of research judgment. It measures reasoning speed and accuracy on problems designed to have one answer. Research at the frontier rewards picking the right problem and persisting with it for years.

Credit in modern AI is also shared. AlexNet had three authors, the translation paper three, the AlphaGo paper many more. Individual scores, even if they existed, would describe only part of what produced those results.

How we assessed this

We report no IQ figure because none has been disclosed or reported by a reputable outlet. Education, dates and honors were checked against encyclopedia records; research results against the original papers on the NeurIPS proceedings site and arXiv. For a sense of where very high scores sit, see IQ 145, the three-standard-deviation mark, and our explainer on what counts as a high IQ. Our practice IQ test is a screener rather than a clinical assessment.

Editorial standard: IQ figures for historical figures are retrospective estimates and are labeled as such; scores for living people appear only when the person disclosed them or a reputable outlet reported them. Corrections: contact the editors.