Reka AI's omni-model Rho-1 handles text, images, video, and robot control in a single model

Most AI systems handle text, images and video with separate models or tools, but Reka AI's Rho-1 feeds every kind of input into one shared context window inside a single 19-billion-parameter network. The research preview also turns the same weights into robot control commands, and its makers trained it on 320 H100 chips over about three months, drawing motion signals from ordinary internet videos.

Summary written with AI by HowTechMoves, based on reporting by The Decoder.