Mental Abacus (MA) is a popular arithmetic technique in which students learn to solve math problems by visualizing a physical abacus structure. Prior studies conducted in Asia have found that MA can lead to exceptional mathematics achievement in highly motivated individuals, and that extensive training over multiple years can also benefit students in standard classroom settings. Here we explored the benefits of shorter-term MA training to typical students in a US school. Specifically, we tested whether MA (1) improves arithmetic performance relative to a standard math curriculum, and (2) leads to changes in spatial working memory, as claimed by several recent reports. To address these questions, we conducted a one-year, classroom-randomized trial of MA instruction. We found that first-graders students struggled to achieve abacus expertise over the course of the year, while second-graders were more successful. Neither age group showed a significant advantage in cognitive abilities or mathematical computation relative to controls, although older children showed some hints of an advantage in learning place-value concepts. Overall, our results suggest caution in the adoption of MA as a short-term educational intervention.