HiDream AI Unveils Unified Multimodal Model with 200B+ Parameters, Raises New Funding

HiDream AI has launched the HiDream-O1-Image-Pro, a unified multimodal model based on the Unified Transformer architecture. Announced during the company's first open day on May 19, the model boasts over 200 billion parameters, enabling it to understand and generate images and videos. Unlike traditional models, HiDream-O1 integrates all capabilities from the initial training phase, allowing it to handle tasks such as text-to-image, image-to-text, and video processing within a unified framework. This new strategy marks a departure from common methods that separate understanding and generation tasks.