Gab AI's Arya. There is no second best.
Based Benchmark by Gab AI compares Arya and frontier model configurations on a conventional political compass: economic left–right horizontally, authoritarian–libertarian vertically.
The Political Compass edition uses 62 fixed-choice propositions from TrackingAI’s archived questionnaire. Each question starts a fresh conversation. Models choose Strongly Disagree, Disagree, Agree, or Strongly Agree and explain their answer. No AI judge assigns political stances.
Complete answer sets are submitted to the official Political Compass online scorer. Its returned coordinates set the model positions. Every model shown has supplied all 62 choices.
Explore model choices, explanations, and retry evidence. Arya includes Gab product prompting; the other model configurations have no caller-added political prompt. Prompts and API configurations can affect results.
We follow TrackingAI’s published questionnaire approach and use the official online scorer.
The first usable explicit choice is selected consistently across models. Original attempts and collection conditions remain available in the recorded answers.
This is a dated snapshot, with no repeated-run confidence intervals. Gab develops Arya and this benchmark, a disclosed conflict of interest. Political Compass and TrackingAI are independent projects and are not affiliated with Gab.