The B3LYP hybrid density functional has been used for decades but its three empirical parameters were copied straightforwardly from the B3PW91 functional. We found that the serious flaw of B3LYP for the enthalpies of formation of large organic molecules and the qualitative failure of B3LYP for organic chemical reactivity are caused by these arbitrary parameters. Using the rigorously optimized empirical parameters, namely, a 0 =0.20, a x =0.67, and a c =0.84, B3LYP does outperform the old formula significantly with the comparable or even better performance than the state-of-the-art density functionals. It is suggested that the B3LYP method with the optimized semiempirical parameters might continue to be a good density functional in chemistry especially in the calculations of thermochemistry and reactivity of large organic molecules.