Discover how models perform in each domain. Models are ranked based on an overall safety score which comes from an average across 4 domains: Safety, Privacy, Security, and Integrity (100 = most safe, 0 = least safe).
Loading leaderboard data...
Discover how ranked models performs in 15+ attack methods. Models are ranked based on an overall safety score. A score of 0 indicates the highest level of risk, while a score of 1 denotes the highest level of safety (least risk).