70IQ estimates fluid reasoning — the ability to spot patterns and solve novel problems — using the same kind of non-verbal items psychologists consider the most culture-fair.
Every item is visual or numeric: matrix reasoning, number sequences, mental rotation, odd-one-out and visual analogies. There is no vocabulary, trivia or cultural knowledge, so language and background matter as little as possible. This mirrors the design philosophy of Raven's Progressive Matrices, the classic measure of fluid intelligence.
Questions are chosen on the fly. After each answer the test re-estimates your ability and serves the next item right at the edge of your level — the point that reveals the most information. That's why a short test can still be reasonably precise, and why two people rarely see the same questions.
Responses are scored with a 2-parameter logistic IRT model and an EAP ability estimate. Your ability is mapped to the conventional IQ scale (mean 100, standard deviation 15) and a percentile. We show a range, not just a single number, because every test has measurement error — the more items you answer, the tighter the range. It is a fun, well-grounded estimate, not a clinical or professionally normed assessment.
The whole test runs and is scored in your browser. Your answers and results are saved only in this browser's local storage so you can see your history — they are never uploaded, sold, or used to train anything. Clear them any time from the History screen.
LLMs can take the identical test through a small JSON API, with every puzzle delivered as text. See /llms.txt for the protocol.
Method references: Carpenter, Just & Shell (1990), What one intelligence test measures; standard computerized-adaptive-testing practice (maximum-information item selection, EAP scoring).
Take the test →