5 ms·Yzma = embedding+inference on VLM/LLM/SLM/TLM in pure Go using llama.cpp1 points by deadprogram 11mo ago