On July 13th, SenseTime (00020.HK) officially released and fully open-sourced its SenseNova-Vision, a unified visual large model for understanding and generation, marking a significant upgrade to the visual capabilities within SenseTime's SenseNova large model ecosystem. The core innovation of SenseNova-Vision lies in enabling vision to become a native capability of general-purpose foundational models and fully integrating it into the large model ecosystem, achieving native unification of all classic visual tasks, including object detection, image segmentation, depth prediction, and 3D reconstruction. In multiple evaluations, SenseNova-Vision led in four core visual domains with a single model, even surpassing specialized expert models. Additionally, SenseTime simultaneously open-sourced the SenseNova-Vision Corpus-50M, a visual instruction corpus containing 50 million samples.