ArcGPT is is a large language model that can hold natural conversations with users and provide intelligent archival services such as archival knowledge Q&A, keyword extraction, smart compilation and research, OCR recognition and retention period prediction.. The model was developed under the leadership of the Smart Archives Laboratory of Wuhan University. The team built large-scale archival knowledge datasets of many types, including archival standards, archival policies and regulations, and academic papers on archives, and annotated prompt corpora for archival business scenarios for targeted pre-training. The model achieved the best results in multiple archival evaluation tasks.