Abstract Tropical Pacific sea surface temperatures (SSTs) exhibit a strong background east‐west gradient, which intensified between 1982 and 2024. This trend is not well captured by global climate models, but whether this is due to model bias or internal variability remains unclear. In this study, we introduce a semi‐Lagrangian metric based on the difference between the warmest and coldest 50% of tropical Pacific SSTs (regardless of their location) which increases the signal‐to‐noise ratio, limiting variability more effectively than the fixed east‐west boxes used by previous studies. By implementing this metric, our study shows that consistency between modeled and observed trends is beyond unlikely, becoming virtually impossible. We determine that the biggest contributor to inaccurate model trends is the coldest 5%, corresponding to the coastal upwelling region off Peru and Chile. We also show that an improved model mean state in this region does not necessarily lead to an improved trend.