July 7, 2026 Scaling Rectified Flow Transformers for High-Resolution Image Synthesis (Stable Diffusion 3) Stability AI 的 SD3 论文：改进的 Rectified Flow 时间步采样 + 多模态 MM-DiT 架构，在高分辨率文生图上超越 DALL·E 3 等 SOTA diffusion-model image-generation text-to-image flow-matching transformer scaling
July 7, 2026 HunyuanImage 3.0 Technical Report 腾讯混元原生多模态模型：在自回归框架内统一多模态理解与生成，80B MoE（激活 13B），当前最大最强开源图像生成模型 image-generation text-to-image vision-language multimodal mixture-of-experts autoregressive