DeepSeek v4.1 Flash (518GB, 4-bit) on a 128GB MacBook: 2.7x prefill, 17 tok/s

DeepSeek v4.1 Flash (518 GB, 4‑bit) runs on a 128 GB MacBook, delivering 2.7× faster prefill and 17 tokens per second. To adopt it, first review the README, act