Over the past decades, writing assessment research has been concentrating on alternative methods with a social-oriented view of assessment, including dynamic assessment (DA). Given the lack of research juxtaposing the interventionist and interactionist DA frameworks in the area of narrative writing, this study sought to compare the effectiveness of Brown’s graduated prompts model vs. Poehner’s model in the development of one-paragraph narrative essays in terms of grammatical accuracy. The study followed a quasi-experimental design, with 15 Iranian EFL learners selected via convenient sampling from among the female students of a language institute in Tehran. The participants were then randomly divided into three groups: Interventionist group, in which mediation was based on Brown’s model in the sandwich format; interactionist group, where mediation was done using Poehner’s model in the cake format; and non-dynamic assessment (NDA) control group with no mediation involved. The research consisted of three pilot sessions and eleven sessions as the main phase. To analyze the data, both descriptive and non-parametric inferential statistics were run. The results conceded the superiority of both DA approaches to NDA, whereas no significant difference was observed between the two DA groups in their general performance on narrative tasks. However, the analysis of the number and types of required mediational moves over the DA sessions indicated the superiority of the interactionist model to interventionist framework in the development of grammatical accuracy in narrative paragraphs. The study offers some theoretical and pedagogical repercussions for educators, curriculum designers, and L2 teachers.