At present, the existing English teaching model is not constructed according to the LASeR standard, resulting in the low I/O performance of the English teaching model and the decline of students’ learning efficiency. A multilevel input English teaching model considering diversified training objectives is proposed. Split the semantic features of multilevel input English teaching model information, calculate the support value of data items in the English teaching model resource database, and delete the complex and large amount of resource information. And with the probability vector diagram allocation method to match the autocorrelation characteristics of teaching resources, based on the relevant feature extraction results, complete the efficient mining of multilevel input-oriented English teaching model resources and realize the construction of multilevel input-oriented English teaching model considering diversified training objectives. The experimental results show that the designed model has high I/O performance and effectively improves students’ learning efficiency.