Open Source · MarkTechPost ·
PrismML Releases Bonsai 27B: 1-bit and Ternary Builds of Qwen3.6-27B That Run on Laptops and Phones
PrismML released Bonsai 27B, low-bit versions of Qwen3.6-27B rather than a new pretrained model. The Apache 2.0 release includes ternary weights at 1.71 bits per weight with an ideal 5.9GB size and a binary 1-bit variant, targeting laptops and phones.