A spatial model of the 3D formation drillability field is crucial for drilling optimization in the petroleum area. In this paper, a new spatial modeling method is proposed for the 3D formation drillability field, which has two stages. In the first stage, the number of formation modes is determined according to the formation characteristics and these modes are identified by the fuzzy c-means clustering algorithm. In the second stage, random forest models are built separately for all formation modes. X, Y ground coordinates and depth coordinate are selected as the model inputs while the model output is the formation drillability. After that, these spatial 3D formation models are combined into one spatial 3D formation model for the whole drillability field. The proposed method and four compared methods (Random forest, ScatteredInterpolant, Support vector regression, and Kriging) are cross-validated and tested by using the data from eight drilling wells in the Xujiaweizi area, Northeast China. The results indicate the effectiveness of proposed method in spatial 3D formation drillability modeling.