at 200M tokens for the full AI suite run its not token efficient at all