The upcoming 3G environment is surely of heterogeneous mobile devices and heterogeneous wireless networks. Many applications such as multimedia streaming, online games and peer-to-peer file sharing will be widely used in such an environment in which users can have their own individual choices of devices and services. In particular, users can carry heterogeneous mobile devices with different cache sizes, transmission range, and even transmission latency. These devices not only can get services from different Mobile Support Stations (MSS) (depending on the wireless networks they have chosen), but also can form a mobile peer-to-peer (P2P) network to provide services to each other. For such a complex mobile network environment, an effective cache mechanism that can handle the heterogeneous properties is required. In this paper, we propose a flexible cache scheme which is adaptive to the actual device condition and that of its surrounding environment (containing other mobile peers and MSS). We extend the notion of "caching" into a more ambitious concept of "intelligent cache" adaptive to the pervasive computing environment. The main techniques adopted by our caching mechanism include Adaptive Data Caching and Least Cost Caching Path Generating, which can then be further combined with a Data Replication via Clustering technique. We term such a resulting mechanism as Blended Caching scheme and the services provided by the scheme as Adaptive Caching Services (ACS). To better utilize and exploit our caching scheme, an ACS-oriented network layer is also designed. Simulation experiments and analytical studies are carried out to evaluate the effectiveness of our adaptive cache scheme.