Seismic intensity measures are the core indicators for determining the distribution of seismic risk and the vulnerability of buildings in large-scale zones. As traditional intensity measures, macroscopic and instrumental seismic intensities are widely used worldwide. However, the quantification methods for macroseismic and instrument intensities are limited by the fuzziness of evaluation scales and the uneven distribution of seismic stations, resulting in an urgent need to improve the accuracy of evaluation scales. This study combines Chinese seismic design standards with macroscopic and instrumental intensity scales, and an evaluation method that considers multivariate intensity measures was proposed. Using nonlinear and response spectrum theory, nonlinear dynamic and spectral analyses were conducted on 360,000 accelerations monitored by 12 stations during the 2022 Luding earthquake in Sichuan Province, China. The time history and response spectrum curves were generated by considering the influence of different earthquake directions. An optimized seismic vulnerability level assessment method is proposed. A vulnerability assessment was conducted on four typical buildings (81239,407 m2 and 938,636 m2) subjected to two destructive earthquakes (the Shidian (SD) earthquake on April 10, 2001, and the Ninger (NR) earthquake on June 3, 2007) in Yunnan Province, China. A seismic damage probability matrix and surface model considering multivariate intensity measures were established. Using an updated nonlinear fitting algorithm, a seismic vulnerability curve for four types of buildings was generated on the basis of an optimized seismic vulnerability level. A seismic risk index model considering the coupling effect of adjacent vulnerability levels is proposed via a fuzzy set and probability theory. Combined with four types of structural damage data from two typical destructive earthquakes, seismic risk membership comparison curves considering multivariate intensity measures are generated.