HiDream-O1 Image-Pro, a flagship image generation model, has been launched.

HiDream-O1-Image-Pro, developed by Zhixiang Future, is based on a native UiT full-modal architecture. It maps image pixels, text, and task conditions to a shared token space, abandoning traditional VAEs and separate encoders. It sets state-of-the-art benchmarks in GenEval and DPG, and ranks 8th in the Artificial Analysis arena, surpassing open-source models such as FLUX.2[Dev].