Model IQ Leaderboard — AI Model Dumbing-down Radar

Summary

Continuously monitors capability changes across 17 AI models including Claude, GPT, Gemini, DeepSeek and Kimi, updating every 5 hours. Shows composite IQ, 30-day trends and per-dimension scores for reasoning, coding, tool use and stability, using statistical change-point algorithms (CUSUM + Page-Hinkley) to detect silent "dumbing-down".