Today for AI
HOT RADAR
toolsHEAT 15.1°

AI CLUSTERED EVENT · 10/13/2026

Janas Open-Source Engine Streams MoE Experts from SSD to Run 48GB Models on 32GB Laptops with 9x Speedup

1 reports archived1 independent sourcesupdated 10/13/2026, 06:05:43
Synthesis & Latest Updates
1 Sources Cross-Validated

Janas releases an open-source LLM inference engine that streams Mixture-of-Experts weights directly from NVMe SSDs, enabling 48GB models to run on laptops with only 32GB RAM. Benchmarks show a generation speed of 24 tok/s, nearly 9 times faster than llama.cpp's 2.8 tok/s, significantly lowering hardware barriers for local deployment.

LATEST/Janas releases an open-source LLM inference engine that streams Mixture-of-Experts weights directly from NVMe SSDs, enabling 48GB models to run on laptops with only 32GB RAM. Benchmarks show a generation speed of 24 tok/s, nearly 9 times faster than llama.cpp's 2.8 tok/s, significantly lowering hardware barriers for local deployment.

HEAT TRENDHourly heat curve

4 fully observed hours
Now
15.1
Peak
15.807:00
24h change
–

The line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.

TIMELINECoverage timeline

Total 1 reports · Latest first
  1. Hacker News AIT2·78 pts
    • Janas uses SSD streaming technology to load and run MoE models larger than physical RAM capacity.
    • On an Intel Core Ultra 9 laptop, Qwen3-Next-80B achieved 24 tok/s generation speed, vastly outperforming llama.cpp's 2.8 tok/s.
    • The project supports various mainstream open-source models (e.g., Qwen, Gemma) and includes benchmarking tools for different hardware environments.