Diffuse Optical Tomography (DOT) is a recently emerging imaging modality that provides optical properties of a tissue. One of main challenges in DOT is the fact that it requires modeling of light propagation in tissue, which is a time consuming and computationally intensive task. However, this process can be accelerated by parallelizing the application. A graphics processing unit (GPU) has the suitable architecture for this task. Therefore, the main goal of this work is to implement a GPU-based solver for the forward problem of DOT. A Finite Element method is utilized to solve the diffusion approximation of the photon transport model in the project. CUDA's parallel architecture and MATLAB software are combined for the implementation of GPU-based forward solver. Several simulations are performed to test computational accuracy and efficiency of the solver. The results show that GPU-based implementation provides a significant speed-up with high accuracy.