WeChat (Tencent), Canmiao Fu et al. · Unified Image Model

WeLM-Image

WeLM-Image is a unified image foundation model for text-to-image generation, instruction-guided editing and multi-reference synthesis, designed for single-GPU deployment in WeChat. It uses a Mixture-of-Transformers backbone with Mixture-of-Experts modules, giving 31.7B stored parameters and about 5.4B generation parameters active per target token. The authors report that the 1536-resolution model ranks first among evaluated open-source models on PRISM-Bench, Qwen-Image-Bench and UnicBench. A MeanFlow-distilled variant, WeLM-Image-Turbo, generates a 1024×1024 image in 4.98 s on one NVIDIA H20 using eight guidance-free steps. Licence and availability are not stated on the page.