SenseTime Releases SenseNova U1: Native Unified Architecture Marks End of 'Stitching' Era

SenseTime has launched its new model, SenseNova U1, marking a significant shift in multimodal architecture. This model employs a unified architecture that integrates understanding and generation, addressing issues found in traditional stitched systems. The new architecture combines images, text, video, and motion into a single representation space, leading to notable performance improvements across various benchmarks. This innovation allows the model to achieve stronger visual reasoning and more natural multimodal generation.