PrismML, founded by Caltech researchers and led by compression expert Babak Hassibi, aims to make high-performing reasoning models small enough for personal devices. The new Bonsai 2 27B model achieves a 9x to 10x memory reduction compared to the original, matching 98% of its aggregate benchmark scores. The company's first Bonsai model, released in March, has been downloaded over 11 million times. PrismML is backed by investors including Khosla Ventures, Cerberus Capital, and Caltech, with Databricks co-founder Ion Stoica serving as an advisor. The startup plans to apply its compression technique to even larger models in the coming months.