PrismML releases 5.9 GB Bonsai 2 model compressed from Qwen 27B

PrismML on Thursday released Bonsai 2 27B, compressing Alibaba's open-source Qwen3.8 27B to 5.9 GB of memory, a 9x to 10x reduction the company says is small enough for a PC and possibly a high-end smartphone. The Caltech-founded startup said the model matches 98% of Qwen's aggregate benchmark scores, up from 95% for the earlier Bonsai, which it said has been downloaded more than 11 million times. PrismML's method stores weights as +1, -1, or 0 instead of 16 bits. It has raised a $22.25 million seed round from Khosla Ventures, Cerberus Capital, and Caltech. CEO Babak Hassibi said upcoming models will target several-hundred-billion parameters, where he expects compression to retain more intelligence. Advisor Ion Stoica said on-device models can run privately without cloud calls.