An open-source benchmark for comparing AI models across reasoning, coding, instruction following, reliability, speed, and resource requirements. Built to make model evaluation more transparent and reproducible.