Llama-modchip: 284B DeepSeek on 64GB RAM, no GPU Link: https://github.com/synchopate/llama-modchip Discussion: https://news.ycombinator.com/item?id=50411961