A systematic comparative study of theoretical geometry optimization on all of the first and second row gas phase inorganic molecules collected in the 77th ed. CRC Handbook of Chemistry and Physics was carried out. The routinely examined theoretical levels for all of the molecules were the analytical gradient MP2(full),B3LYP and B3PW91 with 6-31G(d), 6-31G(d,p), 6-311G(d,p) and 6-311G(2d,p) basis sets. Additional examined theoretical levels for the diatomic, tri-atomic and most of the poly atomic species were QCISD(T)/6-31G(d,p) and QCISD(T)/6-311G(d,p) numerical optimizations. For the diatomic hydrides, extra numerical QCISD(T)/6-31G(d) and analytical QCISD(full)/6-311G(d,p) optimizations were included. For the poly atomic species, the computationally less expensive BPW91 with three basis sets were also examined. The overall performances of the theoretical methods and the basis sets were analyzed in means of the average deviations (AD) and the average absolute deviations (AAD). It was shown that all of the theoretical methods reproduced the bond angles reasonably well. For the bond lengths at the same basis sets, B3PW91 performed the best, MP2 the next and then B3LYP. 6-311G(d,p) basis sets greatly improved the QCISD(T) results from those of 6-31G(d,p). In all of the theoretical methods examined, BPW91 resulted in the largest AAD errors. The AAD values (in pm) for all of the bond lengths in this study were summarized in Table 7 of the text,e.g.,1.50 for B3PW91/6-311G(2d,p), 1.73 for B3LYP/6-311G(2d,p), 1.75 for MP2(full)/6-311G(2d,p), 1.82 for B3PW91/6 311G(d,p),1.82 for QCISD(T,FC)/6 311G(d,p),1.84 for B3PW91/6 31G(d),1.87 for MP2(full)/6 311G(d,p),2.09 for MP2(full)/6 31G(d,p),2.11 for B3LYP/6 311G(d,p),2.12 for B3LYP/6-31G(d), 2.14 for QCISD(T,FC)/6-31G(d,p), 2.88 for BPW91/6-311G(2d,p), 3.33 for BPW91/6-31G(d) and 3.42 for BPW91/6-311G(d,p), with the number of bond length samples of 248, 248, 246, 248, 187, 248, 246, 246, 248, 248, 187, 151, 151 and 151,respectively.