Cross-domain morphing aircraft can change their configurations to adapt to different flight conditions and improve flight capability in large airspace within wide speed range， consequently becoming a topic of interest. This paper constructs a high-fidelity aerodynamic and aerothermodynamic model of a quasi-waverider cross-domain morphing aircraft based on RANS equations and low-fidelity models based on engineering estimation methods. The Non-hierarchical Multi-model Fusion Method using Multi-level Kriging and Quadratic Programming （NMF-MKQP） is proposed considering the existence of non-hierarchical multi-fidelity models for hypersonic aerodynamic and aerothermodynamic analysis. The uncoupled expression of the mean squared error is derived to convert the global optimization problem of scaling factor maximum likelihood estimation into a quadratic programming problem， and the scaling factors are analytically determined accordingly. In this way， the efficiency of the proposed model reduction method is significantly improved while reducing the computation cost. The NMF-MKQP outperforms the state-of-the-art multi-fidelity surrogate modeling methods in terms of approximation accuracy. The aerodynamic and aerothermodynamic characteristics of cross-domain morphing aircraft with varying sweep angles and spans are assessed based on the constructed reduced-order model， further establishing a morphing strategy during cross-domain gliding. The flight performance increases with the lift-to-drag ratio， while ensuring a decreasing maximum heat flux.