This work presents a GPU-accelerated, fully matrix-free topology optimization framework based on Smoothed Particle Hydrodynamics (SPH), highlighting recent advances in mesh-free and particle methods for multi-objective thermoelastic design. The framework is motivated by lightweight thermal systems, in which mass-constrained designs inherently couple thermal transport behavior with thermoelastic response. In such systems, mass reduction leads to localized heat conduction paths and non-uniform temperature fields, which in turn generate thermoelastic stresses, motivating a coupled thermoelastic optimization formulation. A density-based topology optimization scheme is adopted, in which Young’s modulus, thermal conductivity, and the thermal-stress coefficient are interpolated using the Rational Approximation of Material Properties (RAMP) model [1]. The governing equations for steady-state heat conduction and linear thermoelastic equilibrium are discretized consistently within the SPH formalism. Owing to its mesh-free nature, SPH enables flexible representation of complex and evolving design spaces without remeshing. However, the resulting increase in computational workload necessitates efficient parallel solution strategies. In this regard, the localized, particle-based interactions inherent to SPH are well suited for GPU acceleration. GPU-accelerated topology optimization has also been shown to effectively handle large-scale problems, predominantly in finite element-based settings [2]. Accordingly, both state and adjoint equations are solved using a GPU-parallelized, matrix-free preconditioned conjugate gradient (PCG) solver, avoiding global matrix assembly and enabling efficient large-scale computation. The framework is implemented in an in-house CUDA-based SPH code, SOPHIA, and validated through three-dimensional benchmark problems for thermal resistance minimization and structural compliance optimization [3]. Numerical results demonstrate that the proposed multi-objective formulation achieves physically consistent trade-offs between thermal and thermoelastic objectives under mass constraints. Compared with single-objective designs, the additional trade-off in thermal performance is modest, while the reduction in thermoelastic compliance is substantial, resulting in a favorable overall performance gain.