RIG (Reasoning and Imagination in Generalist Policy) is a new approach that combines reasoning and imagination in an end-to-end agent.RIG utilizes a data pipeline to integrate and enrich the content of imagination and reasoning in agent trajectories.The joint learning of reasoning and next image generation in RIG leads to significant sample efficiency improvements and generalization.The synergy of reasoning and imagination in RIG enhances the robustness, generalization, interoperability, and overall performance of the agent.