4 ms·

TBH, if they give us first class support for hardware optimized local LLMs and other "AI", I'll forgive them.