Baidu Wenxin
Wenxin became known as a Chinese chat assistant and now splits into base models, a coding assistant, image generation and document parsing. ERNIE 5.0 moves to native omni-modal modelling while ERNIE-4.5-VL and PaddleOCR-VL open their weights under Apache-2.0.
ERNIE 5.0 is still in preview, but the layering is clear: text, image, audio and video enter one model with context up to 128K. ERNIE-4.5-VL-28B-A3B-Thinking spends 28B total and 3B active parameters on thinking with images and parsing tables, commercially licensed. PaddleOCR-VL compresses a 0.9B vision-language model into document parsing for 109 languages, while Comate, Yige and Wenxin Yiyan sit on top.
