Abstract This paper presents a novel artificial bee colony algorithm for binary optimization in general. Our proposal, named NBABC, features a mechanism which limits the number of dimensions that can be changed in the employed and onlookers bees’ phase. We compare the NBABC to other five binary variants of the continuous ABC, including the state-of-the-art versions for binary optimization, and other four well-known methods. We employ different types of tasks to assess the performance of all the algorithms including the OneMax problem, five variations of the 0/1 Knapsack problems and Feature Selection using eight public datasets. The experiments show that the new proposal obtained competitive results, and in some cases outperformed not only the binary-based ABCs but also the other binary swarm-based and evolutionary-based optimizers.