Multifocal structured illumination microscopy (MSIM) can achieve optically sectioned images with twice the diffraction limited resolution at an imaging speed of 1 Hz and an imaging depth of up to 50 μm. Compared with the traditional wide-field SIM, the MSIM has greater imaging depth and optical sectionning ability, and it is more suitable for long-term three-dimensional (3D) super-resolution imaging of living thick samples. However, the MSIM has some problems, such as slow imaging speed and complex image post-processing process. In this work, a fast super-resolution imaging method and system based on the flat-field multiplexed MSIM (FM-MSIM) is proposed. By inserting a beam shaping device into the illumination light path, the Gaussian beam is reshaped into a uniform flat-top profile, thereby improving the intensity uniformity of excitation multi-spot focal array and expanding the field of view. By elongating each diffraction limited excitation focal point four times along the Y direction to form a new multiplexed multifocal array pattern, the number of scanning steps is reduced, the energy utilization is improved, and then the imaging speed and signal-to-noise ratio are improved. Combined with the sparse Bayesian learning image reconstruction algorithm based on multiple measurement vector model, the image reconstruction steps are simplified, the imaging speed can be improved at least 4 times while ensuring the spatial resolution of MSIM. On this basis, the established FM-MSIM system is used to carry out the super-resolution imaging experiments on the BSC cell microtubule samples and mouse kidney slices. The experimental results prove the fast three-dimensional super-resolution imaging ability of the system, which is of great significance in developing the fast MSIM.