Tencent releases Hy Image3.5 preview image model

AshutoshShrivastava @ai_for_successOriginally posted 2026.09.23

Open original on X

Tencent just released Hy Image3.5 preview, their new image generation model available inside Miora. Compared to previous iterations, the main improvements here focus on physical realism, complex prompt following, and dense typography across both English and Chinese. I tested it across a few completely different styles to see how it holds up. Starting with macro wildlife photography: every individual scale and water droplet renders with authentic depth of field instead of typical AI plastic smoothing. 1/7