PrismML Shrinks 27B AI Model to 5.9GB, Keeps 98% Performance

PrismML released Bonsai 2 27B, a compressed version of Alibaba's open-source Qwen3.8 27B. It uses ternary quantization, storing weights as only +1, -1 or 0 instead of 16-bit numbers, cutting memory use 9x to 10x to 5.9GB. PrismML reports 98% benchmark scores, up from 95% for its March Bonsai, downloaded 11 million times.