Reka releases Rho-1, a 19B model for text, images, video and robot actions

RuntimeWire reports that Reka released Rho-1 as a research preview on October 5. The 19B model combines text, image and video generation with robot actions. RuntimeWire says it was trained on 320 H100 GPUs over three months, and its demo ran in LIBERO simulation rather than on a physical robot. The model has acknowledged limits in grounding, editing and resolution.