The recently established AWI Climate Model (AWI-CM), a coupled configuration of the Finite Element Sea Ice-Ocean Model (FESOM) with the atmospheric model ECHAM6, uses a novel multi-resolution approach: Its ocean component builds on a finite element dynamical core supporting unstructured triangular surface grids, allowing to distribute the grid points in a flexible manner. This allows to concentrate resolution in dynamically important regions, with a continuous transition zone to the coarser resolution in other areas. The model is an ideal tool to study the influence of explicit resolution of smaller scales in dedicated experiments. The unique – spatially seamless – approach might also be of benefit when it comes to temporally seamless prediction, bridging the gap between numerical weather prediction and climate models. A first benchmark set-up of AWI-CM with moderate resolution in the atmosphere (T63) and 25km in key ocean areas, e.g. around the equator, achieved a similar overall simulation performance in a long control simulation compared to well-established CMIP5 models. In particular, the (isotropically) increased equatorial resolution considerably increased the realism of TIW activity and ENSO-related variability compared to standard resolutions. The potential of AWI-CM is further exploited within the EU project PRIMAVERA in the HighResMIP of CMIP6, where we plan to contribute simulations with eddy-resolving resolutions (1/12° or 9-10 km) in key areas of the global ocean, such as the Gulf Stream-North Atlantic Current region, the Agulhas retroflection zone, or the Arctic basin. First simulations show distinct improvements with respect to the development of deep temperature and salinity biases in the North Atlantic Ocean and an overall improvement of surface biases. At even higher resolutions of 4.5 km locally in the Arctic, linear kinematic features emerge in the simulated sea ice distribution with potentially strong impacts on air-sea fluxes in the coupled system. Although the tested set-ups are computationally very demanding (with numbers of grid points comparable to a regular 0.25° grid), the throughput is high at about 8 simulated years per day because of high scalability. In addition, we are about to finish the development of a finite volume version of the ocean model code (FESOM 2). It is already faster than the original FESOM version by a factor of two to three, which will further enlarge the set of computationally feasible applications.