PrismML releases Bonsai 2 27B model compressed to fit on PCs
PrismML released its new Bonsai 2 27B artificial intelligence model on Thursday.
Model compression reduces memory use
The startup compressed the Qwen3.8 27B open source model down to 5.9 GB. This reduction represents a nine to ten times decrease in memory compared to the original version. The resulting file size is small enough to run on personal computers and possibly high-end smartphones.
Company background and funding
PrismML was founded by a group of researchers from Caltech and is led by CEO Babak Hassibi. Investors Khosla Ventures, Cerberus Capital, and Caltech provided a seed round of $22.25 million to the company.
Industry connections and advice
Ion Stoica serves as an adviser for the startup while also directing Berkeley's Sky Computing Lab. The company is not alone in this field, as other firms like Multiverse Computing are also working on large language model compression technology.
Reported by one outlet
Only one outlet has published this. Nothing here has been checked against a second report, so read it as that outlet's account and follow the link below for the original.
Reported by
1 independent outlet. Headline as published. Links open the original report.