Baidu Wenxin multimodal thinking model ERNIE-4.5-VL-28B-A3B-Thinking open source

On November 11th, Baidu officially released the multimodal thinking model ERNIE-4.5-VL-28B-A3B-Thinking. This model has only 3B activation parameters. In addition, Baidu has introduced the innovative ability of "image thinking", allowing this model to have the capability of calling image enlargement and image search tools.