PrismML Brings Tiny LLMs to Smart Glasses
PrismML has adapted a compact vision-language model to Qualcomm's smart-glasses platform, targeting private, low-latency AI that runs locally on wearable hardware.
PrismML is bringing compact multimodal models directly onto smart glasses, testing whether wearable AI can work without constant cloud connectivity.
What happened
The company adapted its 1-bit Bonsai language model to Qualcomm's Snapdragon AR1 Gen 1 smart-glasses platform. The 2B-parameter vision-language model is designed to run locally and answer questions about what the wearer is seeing.
No commercial glasses using the model have yet been announced.
Why it matters
Smart glasses operate under severe limits on battery life, heat, connectivity and processing power. Sending every interaction to the cloud increases latency and energy use while also creating privacy concerns.
Smaller models that can run locally could make wearable AI faster and more reliable while keeping more sensor data on the device.
The bigger picture
The AI hardware race is increasingly about model efficiency, not only model scale. Phones, glasses, robots and other edge devices need specialised models that fit within tight power and memory budgets. That creates room for a different class of AI company focused on compression, architecture and local inference.