According to data released by OpenAI, the new model o3 has a rating of 2727 on Codeforces, which is higher than 99.8% of programmers in the world. All good.

If converted into an IQ of about 157, Einstein in history could still compete with AI. He has not officially tested his IQ, but it is generally believed that his IQ is above 160.

Although AI does not yet have emotional intelligence, as far as IQ is concerned, humans are currently unable to see AI’s taillights.

In fact, when OpenAI released GPT-4o this year, its IQ was only 115, which was within the normal range. The o1 preview model has reached 123, which is equivalent to a doctorate level. Starting from the full version of o1, it reached an astonishing 135, and o3mini broke through the 140 mark.

In just over 7 months, OpenAI improved AI’s IQ by a full 42 points.

If we look at the IQ evolution theory of the Flynn Effect, if humans want to increase their IQ by 42 points, it will take about 140 years to complete.

The Flynn Effect was first systematically proposed in 1983. It mainly reveals the changing patterns of intelligence test scores and also helps people rethink the nature of intelligence and its influencing factors.

The survey shows that there is consistency among different populations and age groups. Whether in developed or developing countries, individuals of different races and ages have observed an increase in IQ. The main factors affecting IQ growth include , improved nutrition, improved education quality, improved health conditions, reduced family size, increased cognitive stimulation and other social reasons.

On average, human IQ generally increases by about 3 points every 10 years.

Another way to demonstrate high IQ is to do math test questions. According to test data released by OpenAI, the o3 model achieved a terrifying 96.7 score in the American AIME 2024 mathematics competition.

Significantly surpasses the preview o1 model, almost killing GPT-4o’s 13.4 in an instant. The o3 model is equivalent to the world's top mathematicians.

Many netizens were quite shocked when they saw the o3 model IQ table, feeling that their IQ was about to be laughed at by the AI.

OpenAI is going to launch an Einstein brain project.

The o1 model scored 133 points in the Mensa test. So, I think this IQ estimation table is absolutely correct!

Importantly, the inference of the o3 model is very beneficial for synthetic learning data. But when it comes to AGI, this has also been described as the ‘singularity’, a point in time when technological growth becomes uncontrollable.

Einstein’s IQ was only 160~

157 IQ may be the highest in human historyThe smartest man. If it's 200+, you have to play another table~

After the release of o3's benchmark score, people's fears are real. But with the AI ​​revolution likely to result in a data shortage over the next four years, what we can achieve in the next year looks less certain. However, human capabilities seem to have been surpassed by AI.

Currently, the o3 model is still in the safety testing stage. I can’t imagine how great it will be after the full public beta. If everyone has an Einstein, the jobs of various tutors may not be guaranteed. At least the competition will be better than it is now. Much more intense. At the same time, many mathematical and physical problems that have troubled mankind for decades or even hundreds of years may be solved.