Janas: Stream large language models from SSD to run bigger models on laptops

Janas is a new method for running large language models that exceed available RAM by streaming model components from an SSD or storage device. The approach achieves 37 tokens per second when running Qwen3-Next-80B, a 48 GB model, on a laptop with only 32 GB of memory.

Benchmarks show Janas outperforms llama.cpp across 13 model comparisons at Q4_K_M quantization levels, with both inference and reading speeds substantially higher. The technique addresses a core constraint in consumer AI: enabling users to run state-of-the-art large models on standard hardware without cloud access or expensive upgrades.