We propose a new benchmark called Face-Human-Bench for the comprehensive and scientific evaluation of multi-modal assistants' abilities in face and human understanding.