Particle packings are methods used to fill a container with particles. These are used to simulate granular matter, which has various uses. Particle packings seek to be dense, however, particle packings are slow, they don't become completely dense, and most only work in simple containers. Currently, several techniques have been proposed to achieve a dense packing, significantly reducing packing construction time, but little progress has been seen in increasing the density of the packing. Particle packings reach an average maximum density of approximately 70% in rectangular and cylindrical containers, and 60% in arbitrary containers. The density of the packings is also known as compaction or solid fraction. The objective of this work is to make a compact packing that in arbitrary containers reaches between 60% and 70% of compaction. For this, a compact periodic packing of spheres is taken as a basis, which from the use of spheres of the same size achieves the highest compaction, that is, it is the most dense. Searched packing is made following a periodic hexagonal pattern, to this are added three sizes of spheres, which are smaller than the initial size, these spheres go in the empty spaces left by the hexagonal packing. The proposed method, reaches densities in arbitrary containers between 60% and 70% in times less than 5 minutes using a parallel optimization on GPU resource.