Spin state energies of iron complexes are important for biochemical applications such as the catalytic cycle of cytochrome P450. Due to the size of these systems and the presence of iron, accurate computational results can be obtained only with density functional theory (DFT). Validation of exchange-correlation (xc) DFT functionals for predicting the correct spin ground state of iron complexes is a rather unexplored area. In this contribution we report a systematic study on the performance of several xc functionals for seven iron complexes that are experimentally found to have either a low, intermediate, or high spin ground state. Standard xc functionals like LDA, BLYP, and PBE are found to disfavor high spin states, whereas hybrid and some meta-GGA functionals do provide the correct spin ground state for all molecules. Recently improved pure DFT functionals such as Handy's optimized exchange (OPTX) also perform well. The origin for the apparent performance of the DFT functionals has been addressed and seems to be related to the inclusion of fourth-order terms (s(4)) of the dimensionless (or reduced) density gradient s in the exchange functional.