A heterogeneous UAVs swarm can adapt to the complex and changeable battlefield environment and improve the effectiveness of collaborative operations by virtue of their complementary functions and collaborative capabilities. Therefore, it has been widely concerned and developed rapidly by various military powers. How to effectively organize heterogeneous UAVs to achieve good cooperative effects has become the focus of the“decentralized” cooperative research of a heterogeneous UAVs swarm, which leads to the optimization problem of structural stability and capability coordination of heterogeneous UAVs grouping in “decentralized” cooperative research. To solve this problem, this paper proposes a “decentralized” collaborative algorithm for the heterogeneous UAV swarm based on bi-level stable matching. Firstly, the “distributed” cooperative operation mode and matching problem of the heterogeneous UAV swarm are described. Then, a preference computing model is constructed based on the capability cooperation among heterogeneous UAVs to transform the trilateral matching problem of three types (reconnaissance, offensive, command) UAVs into a bi-level stable matching problem and construct the corresponding linear integer programming model. Finally, through three groups of experiments, the effectiveness of the model and method proposed in this paper is verified. Compared with the benchmark algorithm, the model has higher efficiency and better effect, and it can also achieve good matching effect in large-scale examples.