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

SenseTime open-sourced SenseNova U1, a next-generation multimodal model built on a Native Unified Architecture.

The model uses the same visual encoder and tokenizer for both understanding and generation, aiming to unify representation across tasks.

This approach is designed to avoid the limits of stitched-together multimodal systems and improve consistency and reasoning.

The release highlights a broader shift toward more integrated multimodal AI architectures.