Image
Hunyuan Image 3
Use it ↗Hunyuan Image 3 by Tencent is an 80B parameter text-to-image model featuring a Mixture-of-Experts (MoE) architecture for high-resolution visual synthesis.
Hunyuan Image 3 is an open source large-scale multimodal model that prioritizes detailed text rendering and complex prompt following. It is specialized for professional design, marketing assets, and infographics, offering industry-leading accuracy for both Chinese and English text. Its architecture consists of a Mixture-of-Experts (MoE) design with 64 specialized modules, activating approximately 13 billion parameters per token for efficient inference. This makes it particularly effective for rendering intricate details and cultural contexts that smaller models may overlook. Native support for resolutions up to 2048x2048 ensures output remains sharp across varied aspect ratios.