The AI Lab PrismML, founded by researchers from the California Institute of Technology (Caltech) and advised by UC Berkeley’s renowned computer scientist Ion Stoica, has developed a specialized version of its small language models for use in smart glasses powered by Qualcomm’s Snapdragon chips. At Qualcomm’s recent Snapdragon Summit, the chipmaker demonstrated PrismML’s 1-bit Bonsai LLM, a lightweight large language model (LLM) that can run directly on smart glasses built using the Snapdragon AR1 Gen 1 Platform. This model is a significant reduction in size—shrunk by four times—compared to larger models, while still maintaining nearly all of their performance on standard AI benchmarks. PrismML’s Bonsai LLM is specifically designed for devices with limited processing power, such as smart glasses. It has 2 billion parameters—essentially the number of variables the model uses to process and understand information—and is optimized for both visual and language tasks. This allows users to point their glasses at an object and ask questions about it in real time, with the model providing instant responses based on what it sees and understands. The broader vision behind PrismML is to develop open-weight AI models—models whose internal structures and weights are freely available—that can run directly on consumer devices. This approach aims to reduce reliance on large, proprietary AI systems that often require significant computational resources and raise privacy concerns due to the need to send data to remote servers. By creating models that can operate efficiently on existing hardware, PrismML hopes to offer a more privacy-conscious and computationally efficient alternative. While the demonstration of the Bonsai LLM on Qualcomm’s platform marks an important step toward this goal, no smart glasses using PrismML’s technology have been officially announced yet. The company is likely still in the process of partnering with hardware manufacturers and preparing the technology for consumer release.