Due to the dynamic nature of Internet of Things (IoT) services hosted by energy-constrained devices, the problem of composing services to provide added-value ones with a reduced energy consumption and a high quality of service (QoS) is attracting more attention. Several existing services composition approaches have limited computation time, QoS utility, and composition lifetime since they do not simultaneously address energy and user’s QoS constraints, consider all of the services during the composition process, or usually require tuning specific algorithm parameters. This paper proposes a group teaching-based energy efficient and QoS-aware services composition approach (GT-EQCA) to deal with the aforementioned limitations. To reduce the composition time, while increasing the composition lifetime and QoS utility, only the relevant services in terms of energy and QoS are considered during the composition process. Furthermore, the composition satisfying the QoS constraints with the highest utility in terms of QoS and energy is determined using the group teaching optimization method, which does not require adjusting specific parameters to achieve satisfactory performance. The large-scale simulation scenarios using a real dataset show that the GT-EQCA approach outperforms four baseline algorithms in terms of composition time, energy consumption, and the QoS utility of the composition.