Janas-LLM runs models larger than RAM by streaming experts from SSD
What happened
On October 13, Hacker News featured Janas-LLM, a scheme for running large language models that exceed a device's memory by reading expert weights from an NVMe SSD on demand and caching the frequently used ones. The example given: running Qwen3-Next-80B-A3B, a 48.4 GB model, on a laptop with 32 GB of RAM. The report covers this loading method and the run itself, with no further performance data.
Written by AI from the coverage · updated 3 hours ago
Coverage
Follow the reports to see the story from different sides.
- Hacker News front pageJanas: Run LLMs larger than RAM by streaming experts from SSD
Janas-LLM 通过从 NVMe SSD 按需读取专家权重并缓存常用专家，在 32 GB 内存笔记本上运行 48.4 GB 的 Qwen3-Next-80B-A3B。
Heat over time
Heat now 9·Comparable peak 9（Oct 13 09:00）·Comparable change over 24 hours –
The trend only compares accounts observed without gaps, so its range may be smaller than the current heat. Hover or tap the chart for each hour; the left and right arrow keys step through it.