A recent state-of-the-art neural open information extraction (OpenIE) system\ngenerates extractions iteratively, requiring repeated encoding of partial\noutputs. This comes at a significant computational cost. On the other hand,\nsequence labeling approaches for OpenIE are much faster, but worse in\nextraction quality. In this paper, we bridge this trade-off by presenting an\niterative labeling-based system that establishes a new state of the art for\nOpenIE, while extracting 10x faster. This is achieved through a novel Iterative\nGrid Labeling (IGL) architecture, which treats OpenIE as a 2-D grid labeling\ntask. We improve its performance further by applying coverage (soft)\nconstraints on the grid at training time.\n Moreover, on observing that the best OpenIE systems falter at handling\ncoordination structures, our OpenIE system also incorporates a new coordination\nanalyzer built with the same IGL architecture. This IGL based coordination\nanalyzer helps our OpenIE system handle complicated coordination structures,\nwhile also establishing a new state of the art on the task of coordination\nanalysis, with a 12.3 pts improvement in F1 over previous analyzers. Our OpenIE\nsystem, OpenIE6, beats the previous systems by as much as 4 pts in F1, while\nbeing much faster.\n