Tropical forest soils contain some of the largest carbon (C) stocks on Earth, yet the effects of warming on the fate of fresh C entering tropical soils are still poorly understood. This research sought to understand how the fate of fresh C entering soils is influenced by warming, soil weathering status, and C chemistry. We hypothesized that compounds that are quickly incorporated into microbial biomass (i.e., greater C use efficiency [CUE]) subsequently have longer-term (255 days) retention in soil. We also hypothesized that relatively weathered soils with greater sorptive capacity also retain more fresh C in the short and longer-terms, and that C in these soils is more resistant to weathering loss compared with less weathered soils. We tested these hypotheses by adding two ¹³C-labeled compounds (glucose and glycine) to three tropical forest soils from a weathering gradient in Hawai'i, and then incubating soils at ambient (16 °C), +5 °C, and +10 °C for 255 days. We found that 255-day ¹³C retention in mineral soil across sites and temperatures was best predicted by two factors: initial retention of ¹³C in mineral soil and initial microbial ¹³CUE (Adjusted R² = 0.78). Carbon compound type influenced ¹³C initial retention, with greater glucose-¹³C retention versus glycine-¹³C retention in mineral soils and microbial biomass, corresponding to greater glucose-¹³C retention in soil at 255 days. Warming had a negative longer-term effect on the retention of ¹³C only in the least-weathered soil, supporting our hypothesis. These results show that initial retention of fresh C in soils via mineral sorption and microbial uptake is a strong predictor of longer-term retention, indicating that immediate C losses are a major hurdle for soil C storage. Also, retention of fresh C appears most sensitive to warming in less-weathered tropical soils, supporting the idea that mineral sorption may provide some protections against warming. Understanding the interaction between soil sorptive properties and warming for C cycling could improve predictions of forest-climate feedbacks for tropical regions.