The protective role of self-compassion in cancer patients' psychological outcomes has been confirmed. However, using a composite score of self-compassion, previous research could not clarify how distinct components of self-compassion may mutually interact. This study, using a person-centred approach, aimed to identify profiles of self-compassion in cancer patients and examined the associations of self-compassion profiles with sociodemographic and medical variables and psychological outcomes.This cross-sectional study included 289 patients with heterogeneous cancer types recruited from two hospitals in Xi'an, China. Latent profile analysis was used to identify distinct profiles of self-compassion. The Bolck-Croon-Hagenaars approach was used to examine how these profiles related to sociodemographic and medical characteristics and psychological outcomes.Five profiles of self-compassion were identified: 'average self-compassion' (54%), 'high self-compassion' (19.4%), 'low self-compassion and low self-coldness' (11.4%), 'high self-compassion and high self-coldness' (8%), and 'average self-compassion and high self-coldness' (7.2%). Patients with the 'high self-compassion' profile tended to be older and report no cancer recurrence, and those with the 'low self-compassion and low self-coldness' profile tended to be female. Patients with the 'high self-compassion' profile reported the fewest depressive and anxiety symptoms while patients with the 'average self-compassion and high self-coldness' profile reported the most depressive and anxiety symptoms.The study revealed five self-compassion profiles in cancer patients, which had different psychological outcomes. Future longitudinal research should investigate the causality between self-compassion profiles and psychological outcomes.