PrismML Shrinks Large Language Models To Run On Local Hardware

By Julie Bort · TechCrunch · 2026-09-17

PrismML Shrinks Large Language Models To Run On Local Hardware

TechCrunch reports that Caltech spin-off PrismML has launched Bonsai 2 27B, a model that compresses Alibaba’s Qwen3.8 down to just 5.9 GB. By utilizing a ternary weight system, the startup allows sophisticated reasoning engines to operate on consumer PCs and smartphones without r

TechCrunch reports that Caltech spin-off PrismML has launched Bonsai 2 27B, a model that compresses Alibaba’s Qwen3.8 down to just 5.9 GB. By utilizing a ternary weight system, the startup allows sophisticated reasoning engines to operate on consumer PCs and smartphones without r