3 ms·

I think you are assuming that inference happens on the same machine/VM that executes code generated by an AI agent.