While previous research on digital multimodal composing (DMC) has examined the efficacy of either an element-based rubric or a process-based model that assesses students’ DMC across stages, the important notion of genre and its value for DMC assessment remains underexplored and undertheorized. Given that as a new literacy practice in L2 writing and related fields, DMC covers a wide range of genres, a genre-based model that incorporates the composing elements and process for assessing DMC is warranted. Driven by theories of genre and multimodality, the study first proposed a multilayered framework that entails DMC structures, functions, modal features and selections. Then the theory-driven framework was tested and modified through collaborative action research with five teachers of a university English for general and academic purposes course in China. The study then drew on student-authored multimodal compositions, interviews and observations with the teachers in order to explore how the teachers assessed DMC, as well as the challenges they encountered. Based on the findings, a refined genre-based model that guides teachers to evaluate DMC as purpose-directed social actions to be constructed with apt multimodal choices within and across four major layers (i.e., base units, layout, navigation, and rhetoric) was developed, with implications discussed.