<p>In this study, the potential predictability (PP) of sea surface temperature (SST) over the tropical Pacific and Tropical Indian Ocean (TIO) in decadal climate prediction project (DCPP) models endorsed by phase 6 of the coupled model intercomparison project (CMIP6) is investigated during boreal summer and winter seasons. PP helps to reveal the upper limit of the prediction skill in an ensemble prediction system. We have employed the information-based measures of mutual information (MI) and relative entropy (RE) to examine the PP of SST. Decadal hindcasts showed skilful PP of SST for lead year-1 (LY-1) during the boreal summer and winter seasons over the Indo-Pacific region. However, models showed higher PP in winter compared to summer SSTs in both the Pacific and TIO regions. The PP of SST declines with lead year more rapidly in the equatorial Pacific region than in the TIO. On the other hand, the PP of TIO SST based on MI showed a stable amplitude/skill across all lead years from LY-1 to LY-10. This indicates that potential predictive skill is higher for TIO SST as compared to equatorial Pacific region. Analysis revealed that the actual prediction skill is lower than the potential prediction skill over the equatorial Pacific region and TIO regions, suggesting the scope for improvements in real-time SST predictions. Further, RE-based predictability information for year-to-year variability shows large RE values over the Niño 3.4 region, in all models, arising from El Niño–Southern Oscillation (ENSO) years, suggesting the importance of ENSO in predicting SSTs in the equatorial Pacific region. Conversely, high values of RE over TIO on the interannual time scale are not associated with ENSO years. The 11-year running correlation between Niño 3.4 and TIO SST PP is unstable over time, showing a negative relationship in some decades and a positive relationship in others.</p>