Janas: Run LLMs larger than RAM by streaming experts from SSD

Janas is a new system that enables running large language models (LLMs) larger than available RAM by streaming model experts from an SSD, achieving significantly faster performance than llama.cpp across multiple models. In benchmarks on a 32GB laptop, Janas-LLM reached 37 tokens/second on an 48GB Qwen model compared to llama.cpp's 2.8 tokens/second, with consistent speed advantages demonstrated across various model sizes. The tool is now available for Linux x86-64 and Windows 10 systems.