A specialist in producing embeddings from Chinese text, this model focuses narrowly on representation learning rather than generation. It operates quietly in the background of pipelines, transforming input into dense vectors that downstream tasks can work with. Its open-weight MIT license makes it straightforward to integrate and modify.