YOLOE (Real-Time Seeing Anything) is a highly efficient, unified, and open object detection and segmentation model from Tsinghua University that operates under multiple prompt mechanisms—text prompts, visual inputs, and prompt-free paradigm—with zero inference and transferring overhead compared to closed-set YOLOs. It introduces Re-parameterizable Region-Text Alignment (RepRTA) for text prompts and Semantic-Activated Visual Prompt Encoder (SAVPE) for visual prompts, achieving state-of-the-art open-vocabulary detection and segmentation in real time. Accepted at ICCV 2025, YOLOE surpasses YOLO-Worldv2-S by 3.5 AP with 1.4× inference speedup.