AbstractSelf‐compacting concrete (SCC) is a specialized type of concrete that features excellent fresh properties, enabling it to flow uniformly and compact under its weight without vibration. SCC has been one of the most significant advancements in concrete technology over the past two decades. In efforts to reduce the environmental impact of cement production, a major source of CO2 emissions, silica fume (SF) is often used as a partial replacement for cement. SF‐modified SCC has become a common choice in construction. This study explores the effectiveness of soft computing models in predicting the compressive strength (CS) of SCC modified with varying amounts of silica fume. To achieve this, a comprehensive database was compiled from previous experimental studies, containing 240 data points related to CS. The compressive strength values in the database range from 21.1 to 106.6 MPa. The database includes seven independent variables: cement content (359.0–600.0 kg/m3), water‐to‐binder ratio (0.22–0.51), silica fume content (0.0–150.0 kg/m3), fine aggregate content (680.0–1166.0 kg/m3), coarse aggregate content (595.0–1000.0 kg/m3), superplasticizer content (1.5–15.0 kg/m3), and curing time (1–180 days). Four predictive models were developed based on this database: linear regression (LR), multi‐linear regression (MLR), full‐quadratic (FQ), and M5P‐tree models. The data were split, with two‐thirds used for training (160 data points) and one‐third for testing (80 data points). The performance of each model was evaluated using various statistical metrics, including the coefficient of determination (R2), root mean square error (RMSE), mean absolute error (MAE), objective value (OBJ), scatter index (SI), and a‐20 index. The results revealed that the M5P‐tree model was the most accurate and reliable in predicting the compressive strength of SF‐based SCC across a wide range of strength values. Additionally, sensitivity analysis indicated that curing time had the most significant impact on the mixture's properties.