Targeted therapies are important for invasive breast cancer (IBC) treatment but are generally not standard-of-care in the neoadjuvant setting. To identify therapies with the potential to improve neoadjuvant treatment response, it is essential to develop patient-derived preclinical models that faithfully reflect the diversity of primary IBC subtypes in patients. Here, we collected data of all breast cancer patients (N=1675) diagnosed in the Antoni van Leeuwenhoek hospital (AVL) for a period of three years, while simultaneously receiving tissue samples of 314 patients to establish a collection of mouse-intraductal patient-derived xenograft (MIND-PDX) models fully recapitulating the heterogeneity of primary IBC. Serial transplantation of lesions resulted in the first large-scale cohort of 60 transplantable IBC-MIND models, comprising 31 luminal (i.e., ER+/HER-), 5 HER2+ and 24 TN models, as well as 7 matching PDX organoid (PDXO) models. We show that our IBC-MIND cohort can serve as a platform for preclinical evaluation of experimental neoadjuvant treatments. For triple-negative IBC, we demonstrate that neoadjuvant treatment does not benefit from addition of a PARP inhibitor, while for estrogen receptor (ER) positive IBC the combination of a CDK4/6 inhibitor and fulvestrant improves neoadjuvant treatment response. Our work provides a valuable resource of primary IBC models to study breast cancer biology and develop novel neoadjuvant treatments.