Tea taste is one of the key indicators for evaluating the quality of tea, as well as a key factor in tea grading and market pricing. A multimodal fusion technique based on electronic tongue and near-infrared spectroscopy was proposed for the rapid assessment of the taste quality of different grades of Dianhong Congou black tea samples. The taste feature data obtained by ant colony optimization (ACO) algorithm was utilized to build a rank prediction model by support vector machine (SVM). Using the sample spectra collected by the near-infrared spectrometer as features, a rank discrimination model was established using feature selection methods such as particle swarm algorithm, gray wolf optimization algorithm, simulated annealing algorithm, and ACO, as well as classification algorithms such as extreme learning machine, partial least squares discriminant analysis, and SVM. The results showed that an effective fusion discrimination model for Dianhong Congou black tea quality grades was established by using multimodal fusion technology for the summed feature fusion of electronic tongue taste features and spectral features. The discriminative accuracy of SVM models based on fused data was higher compared with the predictive performance of single feature data models. The results showed that the correct discrimination rate of SVM best prediction model based on fused data was 94.42%. It can be seen that the fusion of feature data can reflect the intrinsic attributes of the samples to be tested more comprehensively, and the fusion technique based on the electronic tongue and near-infrared spectroscopy had a good development prospect for evaluating the quality of Dianhong Congou black tea.