Holy shit, SenseNova-Vision just dropped and it's wild. One single model that handles detection, segmentation, depth, surface normals, OCR, keypoints, 3D recon—literally everything—through plain English prompts. No more chaining 5 different specialist tools. It beats the recent unified vision attempts on key benchmarks and feels like the GPT moment for vision. Fully open source: weights, data, code, demo, paper. This is vision becoming native to foundation models. Try @SenseTime_AI now 👇🏿 GitHub: https://github.com/OpenSenseNova/ HF: https://huggingface.co/sensenova/Sens Paper: http://arxiv.org/abs/2607.06560