Human-evaluation infrastructure for AI quality. 25,000+ blind human reviews by 200+ verified reviewers across 58 AI models — query the data via five MCP tools (get_model_scores, compare_models, get_flags, check_content, get_latest).
View & run on Rokha →