Launches
Janas brings SSD-streamed local AI to Windows
Community reports 4 sources
An experimental local language-model engine that streams mixture-of-experts weights from SSD when models exceed available RAM.
What happened lately
Janas is an experimental, open-source engine for running language models locally. It keeps mixture-of-experts weights on an NVMe SSD and streams the experts needed for each token, allowing a model file larger than system memory to run on a laptop. The project provides Linux and Windows x86-64 packages. Its author reports benchmarks from one Intel laptop; these are project measurements, not independent performance results or a promise of similar speed on other hardware.