5 ms·Depends on how big the NPU is and how much power/memory the inference model needs.by Spellman 9mo agoDepends on how big the NPU is and how much power/memory the inference model needs.