In this paper, we proposed a discrete unified gas kinetic scheme (DUGKS) into the mixed convection issue coupled in a magnetic field and verified the accuracy by comparing with a exist literature. The flow and heat transfer characteristics of Fe3O4-water nanofluid in the mixed convection are investigated using the DUGKS model. The investigation explores the influence of Hartmann number (0.01 ≤ Ha ≤ 100), Rayleigh number (2 × 103 ≤ Ra ≤ 6 × 105), Richardson number (0.5 ≤ Ri ≤ 10), and nanoparticle fraction (0.01≤ϕ ≤ 0.05) on mixed convection heat transfer and flow characteristics. The results reveal that the critical point for the influence of the Hartmann number on the average Nusselt number is 1. When Ha < 1, the variation of Ha has little effect on Nu, but when Ha > 1, the influence of Ha becomes significant. When Ri = 1, the heat convection between the cold wall and the hot wall is strongest, and the Nusselt number at the cold wall reaches its maximum. The changes in velocity and temperature fields under all conditions are effectively captured by the DUGKS method. Therefore, the proposed method exhibits good mesh adaptability and accurately simulates the mixed convection heat transfer problem.