The study of the sky using the most energetic photons plays a crucial role in detecting and exploring high-energy phenomena in the Universe. Observations conducted over recent years with new ground-based and space-borne gamma-ray instruments reveal that the universe is full of extreme accelerators, i.e., objects with surprisingly high efficiency for acceleration of electrons. In particular such an efficient acceleration is observed in gamma-ray-loud binary (GRLB) systems. GRLBs are a newly identified class of X-ray binaries in which interaction of an outflow from the compact object with the wind and radiation emitted by a companion star leads to the production of very-high-energy gamma-ray emission. Only five such systems have been firmly detected as persistent or regularly variable TeV gamma-ray emitters. All GRLBs detected in the TeV energy range contain a hot, young star and exhibit variable or periodic emission at multiple wavelengths across the electromagnetic spectrum. Details of the physical mechanism of the high-energy activity of GRLBs are not clear yet. Broad multi-wavelength observations are crucial to reveal the characteristic energies of the relativistic wind and better understanding of the nature of these sources. It looks quite possible that all these systems can be understood within the “hidden pulsar” model, and the observed differences are due to the different sizes of the system. In my work, I concentrate on the X-ray and gamma-ray emission observed from gamma-ray binaries PSR B1259−63 and LS I +61◦ 303 with Suzaku, XMM-Newton, Swift, Chandra and Fermi observatories. In PSR B1259−63, the compact source is a young 48 ms radio pulsar orbiting Be-type star with period of 3.4 years. During my studies, I have been intensively involved in the analysis of the results of two multi-wavelength campaigns organized in 2010 and 2014 during the periastron passages in this system. These observations reveal complex spectral variability of the source as it passes through the disk of the companion star. In my work, I have tested different theoretical models trying to reproduce the observed behaviour. For LS I +61◦ 303, I have analyzed historical data of Suzaku, Chandra, XMM-Newton, Swift and for the first time demonstrated the variability of the source column density along the orbit.