“OpenAI conducted a custom, on-premise deployment of its O3 reasoning model on the Venato supercomputer at Los Alamos National Labs.”