In this paper we re-visit the observational relation between X-ray luminosity and temperature for high-z galaxy clusters and compare it with the local L_X-T and with theoretical models. To these ends we use a sample of 17 clusters extracted from the Chandra archive supplemented with additional clusters from the literature, either observed by Chandra or XMM-Newton, to form a final sample of 39 high redshift (0.25 < z < 1.3) objects. Different statistical approaches are adopted to analyze the L_X-T relation. The slope of the L_X-T relation of high redshift clusters is steeper than expected from the self-similar model predictions and steeper, even though still compatible within the errors, than the local L_X-T slope. The distant cluster L_X-T relation shows a significant evolution with respect to the local Universe: high-z clusters are more luminous than the local ones by a factor ~2 at any given temperature. The evolution with redshift of the L_X-T relation cannot be described by a single power law nor by the evolution predicted by the self-similar model. We find a strong evolution, similar or stronger than the self-similar model, from z = 0 to z <0.3 followed by a much weaker, if any, evolution at higher redshift. The weaker evolution is compatible with non-gravitational models of structure formation. According to us a statistically significant sample of nearby clusters (z < 0.25) should be observed with the current available X-ray telescopes to completely exclude observational effects due to different generation detectors and to understand this novel result.