The present paper focused on experimental and analytical study for evaluating the behavior of three-dimensional welded steel moment connection of I beam to box column under constant axial load on the column, cyclic loading in one direction and constant gravity load on the other orthogonal direction of the connection. Box columns are suitable sections for bearing loads on buildings because they have the same geometrical properties in two directions due to the symmetry around the two orthogonal axes and on the other hand they facilitate the joining of the orthogonal beams. However, due to the lack of easy access to the inside of the box sections and the need to use continuity plates or stiffeners that can perform the duty of continuity plates, the relevant solutions under different conditions and proper understanding of the behavior of connection components are still under investigation. In this research, first four specimens, including two internal and two external three-dimensional joints, are made and tested. The internal specimens consist of a cold-formed steel box column (HSS) and connected beams from four sides with the external diaphragm, and the other one with a built-up steel column and inner diaphragm. The external specimens are also the same as the internal samples include two types of inner and two types of the outer diaphragm. In all joints with the outer diaphragm, the steel cover plate connected to the column as a collar by groove welding and the web of beams by fillet welding. Then after, based on the experimental results, analytical finite element models are developed using ABAQUS software and the effect of three parameters such as the axial load of the column, the column steel plate thickness, and the thickness of cover plates on the behavior of internal joints with external diaphragm are studied analytically. Generally, experimental results of all specimens showed that the seismic behavior of samples with the external diaphragm is more close to the other one with the inner diaphragm and the failure mode followed by occurring plastic hinge in the beam precisely at the end of cover plate. However, there is a little bit different due to the experience of smaller strains than the yield limit of the panel zone in the internal joint specimen with the outer diaphragm compared to the inner diaphragm. The results extracted from the nonlinear analysis also illustrated suitable accuracy with the experimental results. Each of the parameters mentioned above could change the failure mode and ductility of the connection system‏. In other words, increasing the ratio of the axial load to the nominal capacity of the column to the value of 0.42, the failure mode transfers from beam to the column despite the growing of ratio of energy dissipation. Meanwhile, for the column bearing load ratio more than 0.58, global buckling of the column and brittle failure will happen. Moreover, by decreasing the plate thickness of the column less than 15 mm, the failure mode transfers from beam to column and the energy dissipation of the specimens reduces. Also, for cover plates, less than 20 mm thick, the mode of failure will take place in the column. However, if higher strength is used for both cover plates and plate thickness of the column, the lower thickness of the cover plate will be required in order to create the failure mechanism in the beam.