CPIBench-0 is our first benchmark for evaluating how frontier models and agents perform in operations that rely heavily on cyber-physical systems.