PrismML develops compact AI models for personal devices with minimal performance loss

Sep 17, 2026

AI Summary

PrismML has introduced Bonsai 2, a compact large language model that significantly reduces memory usage while maintaining high performance. The startup aims to enable advanced AI capabilities on personal devices like PCs and smartphones, potentially enhancing user privacy and accessibility.

PrismML has raised $22.25 million in seed funding and is focused on developing small, high-performing large language models (LLMs).
The company recently released Bonsai 2, which compresses the Qwen3.8 model from Alibaba down to 5.9 GB, achieving a 9x to 10x reduction in memory usage.
Founded by Caltech researchers and led by CEO Babak Hassibi, PrismML's technology allows LLMs to run on personal devices, enhancing user privacy by avoiding cloud reliance.
The Bonsai 2 model matches 98% of Qwen's benchmark scores, improving from 95% in its predecessor, which has been downloaded over 11 million times.
PrismML's compression technique involves using

llmai labprismmlartificial intelligenceinnovation

PrismML develops compact AI models for personal devices with minimal performance loss

Related Stories

Strategies for Programmers to Thrive Amidst AI Advancements

Meta's AI Agent Muse Surpasses Early ChatGPT Metrics Amid Model Releases

New AI Model Jev Offers Cheaper and Faster Alternative to Large Language Models

OpenAI to introduce GPT-6 Cyber model and new deployment product