We present simultaneous 0.65–2.5 μm medium resolution (3300 ≤  R λ  ≤ 8100) VLT/X-shooter spectra of the relatively young (150–300 Myr) low-mass (19 ± 5 M Jup ) L–T transition object VHS 1256−1257 b, a known spectroscopic analog of HR8799d. The companion is a prime target for the JWST Early Release Science (ERS) and one of the highest-amplitude variable brown dwarfs known to date. We compare the spectrum to the custom grids of cloudless ATMO models, exploring the atmospheric composition with the Bayesian inference tool ForMoSA . We also reanalyze low-resolution HST/WFC3 1.10–1.67 μm spectra at minimum and maximum variability to contextualize the X-shooter data interpretation. The models reproduce the slope and most molecular absorption from 1.10 to 2.48 μm self-consistently, but they fail to provide a radius and a surface gravity consistent with evolutionary model predictions. They do not reproduce the optical spectrum and the depth of the K I doublets in the J band consistently. We derived T eff = 1380±54 K, log( g ) = 3.97±0.48 dex, M/H = 0.21±0.29, and C/O > 0.63. Our inversion of the HST/WFC3 spectra suggests a relative change of $ 27^{+6}_{-5} $ K of the disk-integrated T eff correlated with the near-infrared brightness. Our data anchor the characterization of that object in the near-infrared and could be used jointly to the ERS mid-infrared data to provide the most detailed characterization of an ultracool dwarf to date.