Abstract: In the trend of shifting from single to collaborative innovation and from closed to open innovation, the differences of creating, accumulating and absorbing knowledge among cities and regions have become crucial for accounting for the heterogeneities of innovation activities. A number of 289 cities at prefecture level and above in China were chosen as the objects of study and co-patent data was used as a proxy for relational data in order to establish a taxonomy on innovation network patterns from dimensions of intra- and inter-city interaction intensity. Based on the taxonomy, the innovation efficiencies of the case cities with different innovation network patterns were measured and compared by DEA approach. Results showed that: (1) Intra-city collaboration intensity were weak as a whole except provincial capitals which possessed most of high education and science resources and old industrial cities which were associated with a low degree of opening, whereas inter-city linkage intensity were characterized by evident spatial heterogeneities with a gradient weakened from the east to the west, and cities with high level of inter-city collaboration intensity were mainly distributed in major urban agglomerations in eastern China, Beijing-Tianjin-Hebei region, the Changjiang River Delta region, the Zhujiang River Delta region and region of West Coast of Taiwan Straits for instance. (2) These cities were classified into four types, namely network city, outward city, local city and isolated city, based on the dimensions of intra- and inter-city network interaction intensity. Interestingly, the spatial distribution of the cities in eastern China were characterized by a combination pattern of concentric circle that network cities centered which were encircled by outward cities, while both local cities and isolated cities were mainly located in the outer circle. (3) Urban innovation efficiencies of the four types of innovation network patterns were significantly different. Network cities at the core position of the innovation network had the highest innovation efficiencies, followed by outward cities and local cities, and isolated cities were the lowest.