We propose a novel dataset of Regesta with the texts they summarize and their apparatus: REVERINO. This dataset contains documents from two existing collections of Regesta, the Monumenta Germaniae Historica of 1883-1894 and the Auvray edition of 1890-1955. Through a pipeline involving Image Segmentation and Optical Character Recognition (OCR) we convert the existing collections, composed of images of the documents pages into machine-friendly textual versions, for a total of 6266 regesta extracted. Finally, we use REVERINO to measure how well current state of the art language models perform in summarizing latin texts, while still coming short of replacing field experts, they show promising zero-shot results.