Frontier multimodal foundation models (Reka Core, Flash, Edge) that understand text, images, video and audio for next-gen agents.