Despite the value of self- and peer-assessments in supporting learning, research reports mixed findings about self- and peer-assessments of writing, raising intriguing queries about the validity and utility of student-led assessment in a writing classroom. Applying a many-facets Rasch measurement approach, this quantitative study investigated 53 EFL Thai undergraduates' rating performances in self- and peer-assessments of writing during an ongoing classroom in a university setting. Main findings revealed that self-assessors tended to rate more leniently and heterogeneously than peer-assessors whose severity levels were far more homogeneous and closer to the level of teacher severity. Across self- and peer-ratings, students could observably distinguish writing performance quality, thereby assigning varying scores across rubric criteria, and lower-ability students were more variable in their levels of severity than those of high ability. The findings highlight the main and interaction effects of student-led assessment type and writing ability level on students' rating performances. That is, students' rating performances vary according to different types of student-led assessment and different levels of ability. More specifically, assessment type exercises a greater bearing on lower-ability raters' rating variability, while ability level exerts a stronger influence on self-raters' rating variability. The findings provide implications for student-led assessment design and research in the classroom context.