The open-source RNA sequence foundation model structRFM, developed by a team at the University of Science and Technology of China, has been published in the journal Nature. The team has made the complete paired dataset, model weights, and source code openly available to researchers worldwide. The model is built upon approximately 21 million computationally annotated sequence–secondary structure paired data entries. Through algorithmic design, it simultaneously pretrains to learn sequence context and structural dependencies. Given only a sequence as input, it can extract multi-dimensional features for various structure and function inference tasks, achieving internationally leading performance across multiple tasks. Ablation experiments verified the contribution of incorporating structural information and masking strategies to improving model performance.