This paper considers parallel Gröbner bases al-<lb>gorithms on distributed memory parallel comput-<lb>ers with multi-core compute nodes. We summa-<lb>rize three different Gröbner bases implementations:<lb>shared memory parallel, pure distributed memory<lb>parallel and distributed memory combined with<lb>shared memory parallelism. The last algorithm,<lb>called distributed hybrid, uses only one control<lb>communication channel between the master node<lb>and the worker nodes and keeps polynomials in<lb>shared memory on a node. The polynomials are<lb>transported asynchronous to the control-flow of the<lb>algorithm in a separate distributed data structure.<lb>The implementation is generic and works for all<lb>implemented (exact) fields. We present new perfor-<lb>mance measurements and discuss the performance<lb>of the algorithms.