Reka AI launches Rho-1, a general-purpose model for text, images, video and robot control

AI startup Reka AI has introduced Rho-1, a 19-billion-parameter general-purpose model designed to handle and generate text, images, video, and robot control actions within a single neural network. According to Reka AI, the model was trained for three months on roughly 320 H100 GPUs, with compute use that the company says is far below that of today’s top-tier models. Rather than sending different tasks to separate specialized systems, Rho-1 runs multimodal workloads as tokens inside one shared context window. The update was cited by Techub News, referencing The Decoder.