This paper deals with the multi-model assembly line balancing problem (MuMALBP) in a reconfigurable environment. The considered line is composed of a fixed number of workstations and can produce different products in batches. Each product requires an appropriate line configuration. Thus, when the product changes, the line has to be reconfigured to satisfy new requirements related to task precedence and cycle time constraints. Reconfiguring the line consists in reassigning certain tasks between the existing workstations. The objective of this paper is to design a line configuration for each product while minimizing the maximum number of task reassignments whatever the sequence of product arrival. To solve this NP-hard problem, a mixed-integer linear program (MILP) is first formulated. Subsequently, a constructive heuristic and a MILP-based heuristic, named Halt-and-Fix, are developed to solve large-size problem instances. All the approaches are tested and approved on a dataset derived from the well-known instances of the assembly line balancing literature. The numerical results show that the Halt-and-Fix heuristic provides a better trade-off between solution quality and CPU time, compared to the constructive heuristic and the MILP formulation. •A reconfigurable multi-model assembly line balancing problem is considered.•The aim is to design a best line configuration for each product.•The objective function minimizes the maximum number of task reassignments.•A MILP formulation is proposed with pre-processing techniques.•Constructive and MILP-based heuristics are developed.•A comparison between the approaches is given on the basis of a benchmark data set.