Background: Old-old age (80+ years) is associated with substantial cognitive decline. In this population, training-induced cognitive plasticity has rarely been studied. While earlier findings on strategy trainings suggested reduced training gains in old-old age, recent results of an extensive process-based working memory (WM) training have been more positive. Objective: Following up on previous research, the present study aimed at examining the effects of a short WM training in old-old adults and the influence of baseline WM capacity on training gains. Methods: A training group (mean age: 86.8 years) and a matched control group (mean age: 87.1 years) participated in the study. The WM training consisted of five tasks that were trained in each of 10 sessions. To evaluate possible transfer effects, executive functions were assessed with two tests before and after training. The training group was divided via median split in high- and low-capacity individuals to determine the influence of baseline WM capacity on training gains. Results: The training group improved in four of the trained tasks (medium-to-large effects). Training gains were significantly larger in the training group than in the control group in only two of those tasks. The training effects were mainly driven by the lowcapacity individuals who improved in all trained tasks. No transfer effects were observed. Conclusions: These positive effects of a short WM training, particularly for low-capacity individuals, emphasize the potential for cognitive plasticity in old-old age. The absence of transfer effects may also point to its limits.