Search
Baidu Wenxin Makes Text-to-Image Model ERNIE-Image Open Source
1 week ago /
Read about 0 minute
Author：小编
On April 15, 2026, Baidu's Wenxin large-scale model team officially announced that the text-to-image model ERNIE-Image would be open source. Boasting a mere 8 billion parameters, this model is built upon a single-stream Diffusion Transformer architecture. This design allows consumer-grade graphics cards equipped with 24GB of memory to generate ultra-realistic and intricate images, on par with those produced by top-tier commercial models. Presently, both the model weights and inference code have been made available on Hugging Face, with support for the ComfyUI Workflow. Additionally, a GGUF quantization solution has been introduced in collaboration with relevant parties.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
