SenseTime secretly develops multimodal model to rival GPT-Image 2

SenseTime is secretly developing a multimodal model called 'U1 Pro' for design tasks, led by co-founder Lin Dahua. It belongs to SenseTime's Daily New model family and is expected to launch internal beta tests in July. Insiders say the model can think like a designer, handling long design cycles from creation to review, and supports 8K output. In tests, its results rival OpenAI's GPT-Image 2, sometimes even better. Industry watchers say design is becoming the next big arena for multimodal AI, after coding. OpenAI may also release a new image model soon, focusing on design. This move shows Chinese AI companies are speeding up in the creative design track.