4 ms·

OpenAI, to their credit, seems to be focusing pretty heavily on token efficiency in GPT 5.5 and beyond.