PrismML brings tiny LLMs to Qualcomm smart glasses

The AI lab PrismML, founded by Caltech researchers and advised by UC Berkeley's Ion Stoica, has developed a specialized version of its compact language models designed for smart glasses powered by Qualcomm Snapdragon processors.

On Wednesday, during Qualcomm's Snapdragon Summit, the chip manufacturer demonstrated PrismML's 1-bit Bonsai LLM, which runs locally on AI smart glasses built using the Snapdragon AR1 Gen 1 Platform.

Stock photo for illustration only, not from the actual event

As previously reported by TechCrunch, PrismML's core achievement lies in its ability to significantly shrink larger models—by a factor of four—while retaining nearly identical performance on standard benchmarks. The smart glasses iteration is a 2-billion-parameter model optimized specifically for vision and language tasks, enabling wearers to ask questions about their immediate surroundings in real time.

Prism's overarching objective is to promote open-weight AI that executes directly on edge devices, maximizing the utility of onboard computing power. The startup pitches this approach as a viable alternative to relying on the privacy assurances of proprietary AI labs and their constant demand for massive computational resources.