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

PrismML has released Bonsai 27B, a set of 1-bit and ternary quantized builds of Qwen3.6-27B designed to run inference on consumer hardware including laptops and smartphones. By reducing weights to 1-bit or ternary precision, the model achieves dramatic reductions in memory footprint and compute requirements without requiring specialized server GPUs. This is a significant milestone for on-device and offline AI deployment, as 27B-parameter models were previously firmly in cloud territory. Developers building privacy-sensitive applications, edge AI products, or offline-capable assistants now have a credible open-weight option at this scale. The release represents a practical inflection point in the democratization of large model inference.