The AI Lab PrismML — established by researchers from Caltech and guided by Ion Stoica of UC Berkeley — has developed a variant of its compact language models designed for smart glasses utilizing Qualcomm’s Snapdragon processors.

On Wednesday, during Qualcomm’s Snapdragon Summit, the chip manufacturer presented PrismML’s 1-bit Bonsai LLM, which can operate locally on AI smart glasses utilizing the Snapdragon AR1 Gen 1 Platform.

As previously noted by TechCrunch, PrismML’s notable achievement is its ability to significantly reduce the size of larger models (in this instance, by 4x), while still maintaining nearly all of their efficacy on conventional benchmarks. The version for smart glasses is a model with 2 billion parameters, optimized for vision and language, enabling users to inquire about their surroundings in real time.

Prism’s broader ambition is to develop open-weight AI that functions on devices and optimally utilizes the computational resources they already possess. The startup promotes this as an alternative to relying on the privacy commitments of proprietary AI labs and their continual demand for increased computing capabilities.

Launching a model for Qualcomm’s chip marks progress toward that aspiration. However, no smart glasses featuring PrismML have been disclosed as of yet.