Translated data: Institutions including Huazhong University of Science and Technology have released a new multimodal large-scale model benchmark, covering five major tasks and 27 datasets. Evaluation results show outstanding performance in tasks such as text recognition and document question answering, but challenges remain in semantic dependency, handwritten text, and multilingual text. The research team has constructed OCRBench to more accurately assess OCR capabilities, providing guidance for the development of multimodal large models. The introduction of OCRBench offers researchers a comprehensive tool, promoting accurate evaluation and improvement in the OCR field for multimodal large models.