One of the most prominent iterative approaches for solving unconstrained optimization problems is the quasi-Newton method. Their fast convergence and exceptional precision distinguish the quasi-Newton methods. We propose a modified secant relation based on a quadratic model to better approximate the objective function’s second curvature. We then provide a new BFGS method for resolving unconstrained optimization problems based on this modified secant relationship. The proposed method uses both gradient and function values, while the usual Secant relation uses only gradient values. Under appropriate conditions, we show that the proposed method is globally convergent without needing any convexity assumption on the objective function. Comparative results show the computational efficiency of the proposed method from the iteration count and function/gradient evaluations perspective.