PrismML brings its tiny LLMs to Qualcomm-powered smart glasses
TechCrunch Julie Bort
PrismML has a tiny AI model ready for Qualcomm smart glasses. It runs on the device itself, and the glasses still haven’t been announced.
Based on reporting by TechCrunch, Julie Bort — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
PrismML has adapted its tiny language models for smart glasses built on Qualcomm chips. The AI lab, founded by Caltech researchers and advised by UC Berkeley’s Ion Stoica, showed off the work on Wednesday at Qualcomm’s Snapdragon Summit.
The chipmaker used PrismML’s 1-bit Bonsai LLM as the demo piece. Qualcomm said the model can run locally on AI smart glasses based on the Snapdragon AR1 Gen 1 Platform, which is the whole point here: the glasses don’t have to send every request off to a server.
PrismML’s pitch is compression without the usual collapse in quality. The company says this version shrinks larger models by 4x while keeping almost all of their benchmark performance, and the smart-glasses version is a 2-billion-parameter model tuned for vision and language.
That means a wearer could ask what they’re looking at and get an answer in real time. It also fits PrismML’s bigger idea: open-weight AI that lives on devices and uses the computing power people already carry around, instead of leaning on proprietary labs and their appetite for more compute. The catch is simple. No smart glasses running PrismML have actually been announced yet.
My take — AI-written commentary, not fact-checked reporting
This is the right direction, because glasses should not need a cloud round-trip just to describe a chair. The industry keeps pretending every new gadget needs a giant server farm behind it, which is a neat way to make “smart” feel expensive and fragile. Open-weight, on-device models are the less glamorous answer—and usually the better one.
Read more about this at: TechCrunch