SenseTime open-sources its native unified multimodal model, SenseNova U1.5-Lite-Preview.

SenseTime has open-sourced its lightweight, native unified multimodal model, SenseNova U1.5-Lite-Preview, based on the NEO-Unify architecture. It uses only 8 B-MoT parameters to connect visual understanding, inference, generation, and editing. The model natively supports 4K image generation and significantly improves local texture, realistic texture, Chinese and English text layout, and adherence to complex visual instructions. The accompanying Prompt Enhance Skill lowers the barrier to entry for writing long-form creative instructions.