Many countries implement micronutrient powder (MNP) programmes to improve the nutritional status of young children. Little is known about the predictors of MNP coverage for different delivery models. We describe MNP coverage of an infant and young child feeding and MNP intervention for children aged 6–23 months comparing two delivery models piloted in rural Nepal: distributing MNPs either by female community health volunteers (FCHVs) or at health facilities (HFs). Cross-sectional household cluster surveys were conducted in four pilot districts among mothers of children 6–23 months after starting MNP distribution. FCHVs in each cluster were also surveyed. We used logistic regression to describe predictors of initial coverage (obtaining a batch of 60 MNP sachets) at 3 months and repeat coverage (≥2 times coverage among eligible children) at 15 months after project launch. At 15 months, initial and repeat coverage were higher in the FCHV model, although no differences were observed at 3 months. Attending an FCHV-led mothers' group meeting where MNP was discussed increased odds of any coverage in both models at 3 months and of repeat coverage in the HF model at 15 months. Perceiving ≥1 positive effects in the child increased odds of repeat coverage in both delivery models. A greater portion of FCHV volunteers from the FCHV model vs. the HF model reported increased burden at 3 and 15 months (not statistically significant). Designing MNP programmes that maximise coverage without overburdening the system can be challenging and more than one delivery model may be needed.