3 ms·

Assuming this is equivalent to parallel sessions, I would hope so, this is like the entire point of vLLM