<p>One of the problems in multi-manned assembly lines is to produce the highest number of workpieces using a fixed number of workstations. At each workstation there is the possibility of having multiple workers simultaneously performing different tasks. This paper studies this class of problem: the multi-manned assembly line balancing problem (MALBP) with the objective of minimising the cycle time. Moreover, task times are not considered deterministic but dependent on the number of workers at the station. By considering dependent task times, it is possible to represent more realistic situations that take into account the possible interferences between workers. In this paper, a mathematical model is presented. Furthermore, two heuristics based on solving a partition problem with constraints are developed: H_GRP and H_MIX. In particular, H_MIX uses a greedy constructive heuristic procedure, named GRDY_DTT, and combines the resolution of a partition problem with the configuration of neighbouring workstations by task swaps and transfers. The computational experiments indicate that H_MIX is the proposal that performs best. Additionally, we show that H_MIX obtains better results than the ones published in the literature for MALBP-2 with deterministic task times.</p>