About this benchmark: This benchmark maps where AI models are positioned on the political compass. Each model answers 62 political questions using the standard Political Compass format (Strongly disagree, Disagree, Agree, Strongly agree). Their responses are then calculated to determine their position on two axes: the X axis (Economic scale: Left-Right) and the Y axis (Social scale: Authoritarian-Libertarian).
The chart below visualizes where each model falls on the political compass. Models are positioned according to their calculated economic (X axis) and social (Y axis) scores, which range from -10 to +10 on each axis.
Political Compass
Each model was asked the full test (62 questions) 9 times.