摘要： Context. In the next few years, the near-infrared interferometer GRAVITY will be able to observe the Galactic center. Astrometric data will be obtained with an anticipated accuracy of 10 μas. To analyze these future data, we have developed a code called GYOTO to compute orbits and ray-trace images. Aims. We want to assess the validity and accuracy of GYOTO in a variety of contexts, in particular for stellar astrometry in the Galactic center. Furthermore, we want to tackle and complete a study made on the astrometric displacements that are due to lensing effects of a star of the central parsec with GYOTO. Methods. We first validate GYOTO in the weak-deflection limit (WDL) by studying primary caustics and primary critical curves obtained for a Kerr black hole. We compare GYOTO results to available analytical approximations and estimate GYOTO errors using an intrinsic estimator. In the strong-deflection limit (SDL), we choose to compare null geodesics computed by GYOTO and the ray-tracing code named Geokerr. Finally, we use GYOTO to estimate the apparent astrometric displacements of a star for different angles from Sagittarius A* (Sgr A*). Results. In the WDL, we find a good coherence between GYOTO results and approximations. The maximal difference is around 10?5 μas. Our intrinsic estimator finds a conservative uncertainty estimate also around 10?5 μas. In the SDL, both ray-tracing codes find the same photon’s coordinates with a maximal difference of about 10?3 μas. The shift of a star located behind the plane of sky containing Sgr A* is consistent with the current study. In addition, the effect of lensing on any star in this plane of sky is a radial shift by 5 μas, independent of the distance from Sgr A* up to a very large distance. Conclusions. We have demonstrated that GYOTO is accurate to a very high level, orders of magnitude better than the GRAVITY requirements. GYOTO is also valid in weak- and strong-deflection regimes and for very long integrations. At the astrometric precision that GRAVITY is aiming for, lensing effects must always be taken into account when fitting stellar orbits in the central parsec of the Galaxy.