Abstract: Phosphorus limits the growth of crops and is easily fixated to red soil; however, reasonable intercropping can promote phosphorus absorption and reduce phosphorus fixation. Studying the effects of maize and soybean intercropping on phosphorus transformation and mobilization in red soil in the southwestern drylands under different phosphorus application levels is of great significance. Based on four consecutive years of field positioning experiments, two planting modes — maize and soybean intercropping and maize monocropping — were set; four phosphorus application levels — no phosphate fertilizer (P0), 60 kg∙hm⁻² of P₂O₅ (P60), 90 kg∙hm⁻² of P₂O₅ (P90), and 120 kg∙hm⁻² of P₂O₅ (P120) — were also implemented. The effects of maize and soybean intercropping on phosphorus fractions in maize rhizosphere soil and the response of soil phosphorus to the phosphorus gradient were studied using modified Hedley phosphorus classification method. The contribution of different phosphorus fractions to the soil phosphorus activation coefficient (PAC) was investigated using a random forest model. Maize and soybean intercropping increased the available phosphorus content and phosphorus availability in red soil under phosphorus fertilization. Compared with maize monocropping, at P0 level, the available phosphorus content of the intercropping maize rhizosphere soil increased significantly by 70.4% (P<0.01). Maize and soybean intercropping greatly promoted the mobilization of phosphorus in red soil and conversion to the active phosphorus pool. At P0 and P90 levels, the soil PAC of intercropping was significantly increased by 87.4% (P<0.05) and 34.6% (P<0.01), respectively, compared with that of monocropping. Intercropping also increased the proportion of active phosphorus pool to total phosphorus by 15.1% averagely. Among them, the Resin-P content in the inorganic active phosphorus component at the P120 level was significantly increased by 53.7% (P<0.05), compared with in monocropping. Furthermore, the NaHCO₃-Po (organic P extracted by sodium bicarbonate) content in the organic active phosphorus pool was significantly increased by 117.0% and 25.6%, at the P0 and P120 levels, respectively (P<0.05). Intercropping reduced the proportion of stable phosphorus pool in red soil by 1.1% of the total phosphorus. At P90 level, the content of Conc.HCl-Pi (inorganic P extracted from concentrated hydrochloric acid) in the stable phosphorus pool was significantly decreased by 40.2% (P<0.01) compared with maize monocropping. The random forest model showed that soil inorganic phosphorus was the main determinant of PAC, and the mean square error of PAC increased by 14.7% when the predicted value of water-soluble inorganic phosphorus (Resin-Pi) was removed. Maize and soybean intercropping significantly increased the available phosphorus content and PAC in maize rhizosphere soil, increased the proportion of active phosphorus pool and moderately stable phosphorus pool, and decreased the proportion of stable phosphorus pool in maize rhizosphere soil. The mobilization effect of maize and soybean intercropping on the phosphorus pool was significant at low and medium phosphorus levels, but not at high phosphorus level, while soil inorganic phosphorus components had a greater effect on PAC. The results showed that maize and soybean intercropping promoted the mobilization of phosphorus and the conversion of phosphorus to the active phosphorus pool in red soil, especially under conditions of medium and low phosphorus application. However, the effect of the intercropping of maize and soybean on the mobilization of phosphorus in red soil was not obvious under the condition of high phosphorus application.