AI research company PrismML has developed a version of its small language model designed to run locally on smart glasses powered by Qualcomm’s Snapdragon chips. 

The company demonstrated its 1 bit Bonsai language model at Qualcomm’s Snapdragon Summit, where it was shown running on devices built on the Snapdragon AR1 Gen 1 Platform.

PrismML was founded by researchers from Caltech and is advised by UC Berkeley professor Ion Stoica. 

The company specializes in reducing the size of large AI models while maintaining much of their performance on standard benchmarks.

Its smart glasses version contains 2 billion parameters and is designed to process both visual and language information, allowing users to ask questions about objects and scenes they are viewing.

The company’s broader goal is to develop open weight AI models that can operate directly on consumer devices rather than relying entirely on cloud based systems. 

PrismML argues that local processing could allow devices to make better use of their available computing power while reducing dependence on proprietary AI services.

The Qualcomm demonstration represents a step toward that goal, although no smart glasses running PrismML’s technology have been formally announced. 

The company’s work highlights the broader push to bring increasingly capable AI models directly onto smaller consumer devices.