PrismML brings its tiny Bonsai LLM to Qualcomm smart glasses chips

PrismML, an AI lab founded by Caltech researchers and advised by UC Berkeley’s Ion Stoica, has built a version of its tiny language models for smart glasses running on Qualcomm’s Snapdragon chips. Qualcomm showcased the model, PrismML’s 1-bit Bonsai LLM, at its Snapdragon Summit on Wednesday; it can run locally on AI smart glasses built on the Snapdragon AR1 Gen 1 Platform.

As TechCrunch has previously reported, PrismML’s claim to fame is shrinking larger models substantially — in this case, by 4x — while retaining almost all of their performance on standard benchmarks. The smart glasses version is a 2-billion-parameter model tuned for vision and language, letting wearers ask what they’re looking at in real time.

PrismML’s larger goal is open-weight AI that runs on-device and makes better use of the computing power devices already have, which the startup pitches as an alternative to relying on the privacy promises of proprietary AI labs and their growing need for more compute. Releasing a model for Qualcomm’s chip is a step toward that vision — though no smart glasses running PrismML have been announced yet.

Why it matters

Shrinking capable models to run entirely on-device, rather than routing queries to the cloud, is the difference between smart glasses that need a constant data connection and ones that work — and keep data local — anywhere.