The paper presents an innovative approach to quickly and efficiently test new nuclear data librairies through the combined use of open source tools (library processing and deterministic transport code) and simple depletion calculations. The use of the open source deterministic transport code Dragon together with WIMS-D Library Update Project (WLUP) and PyNjoy tools allows a fully transparent process from evaluated nuclear data librairies to reactor physics calculations. When compared to a Monte Carlo reference (Serpent2), large discrepancies in terms of k-inf and 239Pu concentrations are obtained with Dragon during pin cell depletion calculations. This is further amplified by the effects of various code-specific options (energy released by fission, resonance upscattering). However, these discrepancies are shown not to affect the conclusions of a comparison between two Dragon calculations performed with the same computational options but rather using two different nuclear data libraries. The computational time is however much reduced allowing brute force sensitivity analysis (one-at-a-time approach). The performance of various recent Joint Evaluated Fission and Fusion (JEFF) evaluated nuclear data libraries (JEFF-3.3, JEFF-4T2.2, and JEFF-4T4) are assessed using this approach. It is demonstrated in the paper that using JEFF-3.3 or JEFF-4T2.2 instead of JEFF-3.1.1 leads to large k-inf differences together with burnup-dependent trends. These issues are mainly due to the nuclear data of major actinides like 235U, 238U, and 239Pu. The issue of excessive reactivity loss with burnup appears to have been resolved in JEFF-4T4.