Reprogramming blood cells to induced pluripotent stem cells (iPSCs) provides a novel tool for modeling blood diseases in vitro. However, the well-known limitations of current reprogramming technologies include low efficiency, slow kinetics, and transgene integration and residual expression. In the present study, we have demonstrated that iPSCs free of transgene and vector sequences could be generated from human BM and CB mononuclear cells using non-integrating episomal vectors. The reprogramming described here is up to 100 times more efficient, occurs 1-3 weeks faster compared with the reprogramming of fibroblasts, and does not require isolation of progenitors or multiple rounds of transfection. Blood-derived iPSC lines lacked rearrangements of IGH and TCR, indicating that their origin is non-B- or non-T-lymphoid cells. When cocultured on OP9, blood-derived iPSCs could be differentiated back to the blood cells, albeit with lower efficiency compared to fibroblast-derived iPSCs. We also generated transgene-free iPSCs from the BM of a patient with chronic myeloid leukemia (CML). CML iPSCs showed a unique complex chromosomal translocation identified in marrow sample while displaying typical embryonic stem cell phenotype and pluripotent differentiation potential. This approach provides an opportunity to explore banked normal and diseased CB and BM samples without the limitations associated with virus-based methods.