Abstract: Knowledge flow in innovation ecosystem has important meaning for coordinative innovation and upgrading the related industrial chain. In the meanwhile, the distribution characteristics of knowledge potential of innovative bodies are key factors which influence the knowledge flow. According to the knowledge stock and knowledge diffusion direction of innovative bodies, high knowledge potential enterprises, medium knowledge potential enterprises and low knowledge potential enterprises are distinguished. By constructing the tripartite evolutionary game model, we study the evolution of knowledge flow behaviors, passive behaviors and opportunism behaviors. The research shows that the potential difference caused by knowledge breadth is conducive to knowledge flow, but potential difference caused by knowledge depth is not conducive to knowledge flow in innovation ecosystem. The absorption & transforming capacity and unobstructed flow is benefiting for evolution of knowledge sharing behaviors. The knowledge flow behaviors have the feature of path dependence, so the passive behaviors and opportunism behaviors should be avoided in the early stages. Because great knowledge potential exists between core enterprises or industry leaders and start-ups or clustered firms, knowledge transferring and absorption therefore becomes difficult. In that case, medium knowledge potential enterprises act both as intermediary and aiders of knowledge flow.