Cellular heterogeneity plays a key role in the development, evolution and metastatic progression of many cancers. Breast cancer has long been classified into a number of molecular subtypes (such as luminal A and B, basal-like, and Her2-enriched) that predict prognostic outcomes and therefore influence clinical treatment decisions. To date, these subtypes have only been described at a bulk tumour level. Advances in single-cell technologies are providing powerful tools for the isolation and molecular profiling of breast cancers at cellular resolution. To explore the cellular heterogeneity of molecular subtype in breast cancer we first used a panel of genes containing the transcriptional markers that define the PAM50 molecular classifier. Five breast cancer cell line models (MCF7, BT474, SKBR3, MDA-MB-231, and MDA-MB-468) were selected as representatives of the molecular subtypes. The Fluidigm C1 and Biomark systems were used to isolate and quantify the gene expression of single cells from each of these models of breast cancer. To quantify the level of cellular heterogeneity within each of these models we applied the PAM50 predictor to isolated single-cells, as well as bulk cell populations. Using this approach we identified clear heterogeneity of molecular subtypes at a single-cell level, where cells with different subtypes exist within a cell line. To extend this approach into more clinically relevant samples we also explored the transcriptional heterogeneity present in cancer cells from patient derived xenograft (PDX) models. Using the same targeted gene expression approach, as well as single-cell RNA-Seq, we again identified significant intra-tumour heterogeneity of molecular subtype as well as sub-populations of cells that showed clinically relevant transcriptional signatures related to cancer stem-cell, basal-like and hypoxic phenotypes. These results suggest a high degree of functional heterogeneity within breast cancer cell lines and PDX models.