Reka Releases Rho-1: A 19B Omni-Reasoning Model That Understands, Generates Video and Outputs Robot Actions in One

Reka has introduced Rho-1, a 19-billion-parameter omni-reasoning model that integrates text, image, video understanding and generation, plus robot action output in a single neural network, eliminating the need for separate modality-specific models. Rho-1 uses a dual-stream transformer architecture with shared attention and a KV cache, enabling efficient multimodal processing and faster video generation, including a distilled variant that significantly speeds up output with minimal quality loss.