Sixth-generation (6G) networks are considered as key enablers in the development of the intelligent information society offering ultra-reliable low-latency communications in the context of massive machine-type communications and the emergence of the Internet-of-everything. In this paper, we consider a short-packet cooperative nonorthogonal multiple access (C-NOMA) communication system that uses the K-means algorithm for user clustering. The short-packet component of the system facilitates low-latency communications. In addition, C-NOMA using decode-and-forward relays allows extended coverage in terms of user equipment (UE) clusters and spectral efficiency in terms of sum rate. Given the trend toward irregular deployment of heterogeneous infrastructure elements in 6G networks and massive amounts of UEs, the K-means algorithm is used to generate clusters of UEs around the locations of the relays that assist the communication between the base station and the UEs. Analytical expressions of the block error rate (BLER) and sum rate are derived allowing a performance assessment of the short-packet C-NOMA communication system. Numerical results are provided revealing the impact of the number of channel uses and the number of UEs on the BLER and sum rate of the considered system. © 2023 IEEE.