Tencent Hunyuan Image 3.0 is the world's largest open-source text-to-image model with 80B total parameters and 13B activated parameters during inference.