A new van der Waals density functional (vdW-DF2) is proposed, improving upon the original vdW-DF by using a more accurate semilocal exchange functional (PW86) and a large-N asymptote gradient correction for the vdW kernel. This method significantly enhances the accuracy of equilibrium separations, hydrogen bond strengths, and van der Waals attractions at intermediate distances. The vdW-DF2 shows excellent agreement with quantum chemical calculations for 22 molecular duplexes, particularly in predicting the correct behavior of van der Waals forces at distances beyond equilibrium. The improved functional addresses the overestimation of equilibrium separations and underestimation of hydrogen bond strengths in the original vdW-DF. The method is implemented in existing codes with minimal modifications and has been tested on extended systems, including graphite interlayer binding and hydrogen adsorption in metal-organic frameworks (MOFs). The results demonstrate that vdW-DF2 provides more accurate predictions for sparse materials and biological systems, with binding energies within 50 meV of reference values. The new functional also improves the prediction of hydrogen-benzene interactions and hydrogen adsorption in MOFs, showing reduced errors compared to the original vdW-DF. Overall, vdW-DF2 represents a significant advancement in the accurate description of noncovalent interactions in various materials.A new van der Waals density functional (vdW-DF2) is proposed, improving upon the original vdW-DF by using a more accurate semilocal exchange functional (PW86) and a large-N asymptote gradient correction for the vdW kernel. This method significantly enhances the accuracy of equilibrium separations, hydrogen bond strengths, and van der Waals attractions at intermediate distances. The vdW-DF2 shows excellent agreement with quantum chemical calculations for 22 molecular duplexes, particularly in predicting the correct behavior of van der Waals forces at distances beyond equilibrium. The improved functional addresses the overestimation of equilibrium separations and underestimation of hydrogen bond strengths in the original vdW-DF. The method is implemented in existing codes with minimal modifications and has been tested on extended systems, including graphite interlayer binding and hydrogen adsorption in metal-organic frameworks (MOFs). The results demonstrate that vdW-DF2 provides more accurate predictions for sparse materials and biological systems, with binding energies within 50 meV of reference values. The new functional also improves the prediction of hydrogen-benzene interactions and hydrogen adsorption in MOFs, showing reduced errors compared to the original vdW-DF. Overall, vdW-DF2 represents a significant advancement in the accurate description of noncovalent interactions in various materials.