This paper has built performance model of knowledge sharing in industry-university cooperation based on Structural Equation Modeling (SEM) to explore the relationship between performance and the main elements, the object elements, the vector elements, and the environmental elements. The conclusion is that the learning performance and system performance in knowledge sharing in industry-university cooperation are greatly affected by the choice and the overflow of their own knowledge, knowledge integration and the formation of common knowledge, the absorption and application of knowledge; knowledge matching and maturity affect on learning performance significantly but on system performance slightly. Among all the elements, the knowledge matching and maturity are influenced by the industry-university partnerships, shared network and platform, shared regulation, cultural and service environment, while their choice and overflow of own knowledge, knowledge integration and the formation of common knowledge and the absorption and application of common knowledge is limited by industry-university partnerships, psychological distance matching degrees, shared networks and platforms, shared regulation, cultural and service environment.