2 ms·

I think this being based on Llama makes it tenable. GPT4 and Gemini are monolithic multipod models. Multiple instances of Llama run per machines.