241,063 people from 196 countries took the same 20-question IQ test. Tap a country to see how it did.
Swipe the map to see every country ↔
In every one of the 53 countries, at least 1 in 28 test-takers scored above the average of #1 Taiwan (116.3). Every country has top minds.
Two random people from the same country differ, on average, by more than the whole gap between #1 and #53 (22.7 points). True in 0 of 53 countries.
people got all 20 questions right, from 77 different countries. Japan has the most perfect scores: 880.
separate Japan, New Zealand and the USA — the closest race in the top 10.
Japan is the highest-ranked country that took the test in its own language (Japanese) rather than English.
people took the test in Indonesia — more than any other country, 32% of everyone on the map.
Across countries with 100+ English-version test-takers (same language, removing translation effects), scores track education and economic output closely — while domestic income inequality (Gini) shows almost no correlation. See all scatter plots & country tables on /method →
All 53 countries with enough data, from #1 to #53. 100 = the average of everyone who took the test. A country joins once 300 people from it have taken the test.
The same 20-question bank, given once to each of 16 AI models. A model's run is a single attempt, not an average of thousands of people like the country figures above — same scale, not a like-for-like comparison. Snapshot 2026-09-25.
Every test adds to your country's average. 20 questions, 5 minutes, free.
Take the free test →How it works: 100 is the average score of everyone who has taken the test since 2026-07-15, and 15 points is one standard deviation. People choose to take the test online, so this shows how our test-takers scored — it is not an official national measurement. Indonesia supplied 32% of all test-takers, so its result weighs heavily on the world average — which is why 35 of the 53 countries sit above 100. Data snapshot 2026-11-02. How this is calculated, a language-translation caveat in the data, and a check against school results and income →