In this paper, a model-free distributed sliding mode control (SMC) method is proposed for the gait balance control problem of bipedal reconfigurable robot. First, the dynamic model of the reconfigurable robot is established via Newton-Euler algorithm. According to the virtual model control (VMC) method, the corresponding virtual forces is obtained to address the desired constrained torques through the Jacobian matrix. In addition, a radial basis function neural network (RBFNN) is utilized to approximate the dynamics of the bipedal reconfigurable robot. Then, an improved model-free distributed coordinated control strategy is designed for gait balance task. On the basis of Lyapunov stability theory, the closed-loop robotic system is proved asymptotically stable. Finally, the effectiveness of the proposed approach is verified by simulation.