Advances in spatial transcriptomics technologies produce RNA imaging data at increasingly higher throughput and scale. Current computational methods identify and measure the spatial relationships between cell-types, but do not leverage the spatial information of individual RNA molecules to reveal subcellular spatiotemporal dynamics of RNA processing. Here, we developed Bento, a computational framework for subcellular analysis of high-throughput spatial transcriptomics datasets. Bento handles single-molecule data generated by diverse spatial transcriptomics technologies and computes spatial statistics of subcellular RNA molecular distributions, compartmental expression, and cell morphology to build multidimensional feature sets for exploratory analysis. We also developed a multi-label ensemble model for generalizable classification of subcellular localization of every gene in every cell. To demonstrate Bento's utility, we applied it to analyze spatial transcriptomics datasets generated by seqFISH+ (10k genes in ~200 fibroblast cells) and MERFISH (130 genes quantified in ~1000 U2-OS cells) to understand the interplay between gene function, cellular morphology and RNA localization. To understand the role of RNA localization in RNA processing, we integrated spatial data with RNA binding protein (RBP) binding data to explore the spatiotemporal dynamics of RBP-RNA interactions at unprecedented scale (3,165 RNA species x 148 RBPs). We found RNA targets of individual RBPs to be enriched in specific subcellular compartments — such as Splicing Factor 3a Protein Complex (SF3A3), and that preferential localization is influenced by RBP binding of genomic regions for RBPs including Staufen homolog 2 (STAU2). Bento builds on the existing ecosystem of single-cell and spatial analysis toolkits, to ensure accessibility and community-driven tool development. We provide Bento as an open-source tool for the community to further expand our understanding of subcellular biology.