RoleLLM is a role-playing framework designed to build and evaluate the role-playing capabilities of large language models. It consists of four stages: role summary construction, context-based instruction generation, role prompting using GPT, and role-based instruction adjustment. Through Context-Instruct and RoleGPT, we created RoleBench, a systematic and fine-grained role-level benchmark dataset containing 168,093 samples. Moreover, RoCIT achieved significant improvements in role-playing ability on RoleBench, producing RoleLLaMA (English) and RoleGLM (Chinese), even comparable to results obtained with GPT-4-based RoleGPT.