PrismML Releases Bonsai 27B: 1-bit and Ternary Builds of Qwen3.6-27B That Run on Laptops and Phones

PrismML has released Bonsai 27B, a compressed version of the Qwen3.6-27B model that utilizes ternary and binary weight representations. By reducing the weight precision to as low as one bit, the model requires significantly less memory, allowing a 27-billion parameter system to operate on devices like laptops and phones.