Abstract: A novel neural network model for solving nonlinear programming problems is proposed in this paper. It is composed of variable neurons, Lagrange multiplier neurons and Kuhn-Tucker multiplier neurons which are interconnected. By making the Kuhn-Tucker multiplier neurons operate in an one-sided saturated mode, the introduction of the slack variables is no more necessary in dealing with the inequality constraints of nonlinear programming problems. This method can avoid the increase in the number of neurons caused by the slack variables. This is advantageous to the hardware implementation and the convergence rate improvement. It can be shown that under suitable conditions the trajectory of the proposed neural network model converges to the equilibrium point corresponding to the optimal solution of the nonlinear programming problem.