We present a joint model for entity-level relation extraction from documents.\nIn contrast to other approaches - which focus on local intra-sentence mention\npairs and thus require annotations on mention level - our model operates on\nentity level. To do so, a multi-task approach is followed that builds upon\ncoreference resolution and gathers relevant signals via multi-instance learning\nwith multi-level representations combining global entity and local mention\ninformation. We achieve state-of-the-art relation extraction results on the\nDocRED dataset and report the first entity-level end-to-end relation extraction\nresults for future reference. Finally, our experimental results suggest that a\njoint approach is on par with task-specific learning, though more efficient due\nto shared parameters and training steps.\n