Multipulse airborne transient electromagnetic (ATEM) systems transmit one high-power pulse and one low-power pulse containing more high-frequency EM signals. Such systems have better near-surface resolutions while maintaining the depth of exploration of other conventional systems. ATEM systems are especially suitable for geologic mapping and mineral exploration. The inversion of multipulse ATEM data has been mainly limited to 1D modeling, which is not suitable for complex underground structures. We have investigated an algorithm for 3D multipulse ATEM data inversion based on direct Gauss-Newton optimization with quite-fast convergence. The forward problems were solved in the frequency-domain based on the secondary scattered electrical field equation, and then the inverse Fourier transform and the convolution with transmitting waveform were applied to calculate the arbitrary waveform response and sensitivity matrix in the time domain. To optimize the number of computations and memory, we further used an EM “footprint” concept in our inversions to reduce the forward model size and sparse the sensitivity matrix. The inversion results of synthetic data showed that our 3D algorithm is very effective for inverting the multipulse data with results combining advantageous resolutions of different transmitting pulses. Finally, we applied our algorithm to invert real survey data obtained at McMurray, Alberta, Canada, to further test its effectiveness.