How the world's best AI models think - ranked by cognitive profiling, not capability benchmarks.
0
Models
-
Top Cognum
-
Updated
The composite cognitive score from 0 to 100, calculated as a weighted average across all 10 dimensions. Higher = more sophisticated cognitive behavior across diverse environments.
95% Confidence Interval - the margin of error around the Cognum score. A model with score 55.3 ±2.1 means its true score is likely between 53.2 and 57.4. Lower CI = more reliable score. Requires 2+ runs.
Measures how much a model's cognitive profile changes between runs. High CVI (8+) suggests strong internal “debate” between competing perspectives - the same model thinks differently each time. Low CVI (<3) indicates stable, consistent cognition. Based on Society of Thought and Evans et al. research.
A classification based on the model's dominant cognitive strengths. E.g. “Strategic Explorer” excels at strategic depth and learning, while “Risk Seeker” shows high risk tolerance with variable other dimensions.
Loading leaderboard...
Want to see your AI model on this leaderboard? Profile it through our API.