The duality theory is well-developed for non linear programs. Technically, a major part of such broad framework possibly be extended to mixed non linear programs, however this has demonstrated complicated, in minority as the duality theory does not integrate well with modern computational practice. In this paper, we constructed a new pair of second-order multiobjective mixed dual problems over arbitrary cones with multiple arguments, with an eye towards developing a more practical framework. Weak, strong and converse duality theorems are then established under $K-\eta$-bonvexity assumptions. Several known results are obtained as special cases.