The COVID-19 pandemic had posed a heavy burden to the healthcare system worldwide. Despite many deep learning models have been proposed to conduct COVID-19 clinical predictive tasks in intensive care, we are the first to provide a comprehensive benchmark for a fair comparison. We propose the Outcome-specific length-of-stay and Early mortality prediction task for patients in intensive care units. Our benchmark and online platform can further facilitate healthcare researchers and clinicians for the COVID-19 predictive modeling.