In the realm of cross-modal Image-Text Retrieval (ITR), this paper focuses on the effectiveness in the Chinese language environment. Based on Chinese CLIP, the model pretrains on a large-scale Chinese image-text dataset and employs carefully selected vision and text encoders with a two-stage pretraining strategy, which enables the model to develop a nuanced semantic alignment between images and texts within a Chinese context. This approach facilitates a deep understanding of the matching between images and Chinese texts. To further enhance the model’s capabilities, the paper introduces domain-specific dataset to implement fine-tuning strategies. Well-designed experiments are conducted between the popular models and ours on public dataset Flickr30K-CN, COCO-CN, and domain-specific dataset AP-ID. Our model attains state-of-the-art (SOTA) results on Chinese text-image retrieval datasets, demonstrating its robustness and effectiveness in handling Chinese-language cross-modal ITR task.