Overexpression of human epidermal growth factor receptor 3 (HER3) is involved in resistance to several therapies for malignant tumours. Currently, several anti-HER3 monoclonal antibodies are under clinical development. We introduce an alternative approach to HER3-targeted therapy based on engineered scaffold proteins, i.e. affibody molecules. We designed a small construct (22.5 kDa, denoted 3A3), consisting of two high-affinity anti-HER3 affibody molecules flanking an albumin-binding domain ABD, which was introduced for prolonged residence in circulation. In vitro, 3A3 efficiently inhibited growth of HER3-expressing BxPC-3 cells. Biodistribution in mice was measured using 3A3 that was site-specifically labelled with 111In via a DOTA chelator. The residence time of 111In-DOTA-3A3 in blood was extended when compared with the monomeric affibody molecule. 111In-DOTA-3A3 accumulated specifically in HER3-expressing BxPC-3 xenografts in mice. However, 111In-DOTA-3A3 cleared more rapidly from blood than a size-matched control construct 111In-DOTA-TAT, most likely due to sequestering of 3A3 by mErbB3, the murine counterpart of HER3. Repeated dosing and increase of injected protein dose decreased uptake of 111In-DOTA-3A3 in mErbB3-expressing tissues. Encouragingly, growth of BxPC-3 xenografts in mice was delayed in an experimental (pilot-scale) therapy study using 3A3. We conclude that the 3A3 affibody format seems promising for treatment of HER3-overexpressing tumours.