{
  "id": 9609813,
  "title": "PrismML brings its tiny LLMs to Qualcomm-powered smart glasses",
  "url": "https://urgent.news/2026/09/24/prismml-brings-its-tiny-llms-to-qualcomm-powered-smart-glasses",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-24T19:00:42.000Z",
  "source": {
    "name": "TechCrunch",
    "slug": "techcrunch",
    "url": "https://techcrunch.com/2026/09/24/prismml-brings-its-tiny-llms-to-qualcomm-powered-smart-glasses/"
  },
  "original_language": "en",
  "account": "PrismML, an AI lab founded by researchers from Caltech and advised by UC Berkeley's Ion Stoica, has developed a compact language model suited for smart glasses powered by Qualcomm's Snapdragon chips. This model, called the 1-bit Bonsai LLM, is showcased at Qualcomm's Snapdragon Summit. This is a significant step for PrismML, which has gained recognition for its ability to shrink larger models by a factor of four while maintaining their performance on standard benchmarks. This startup's primary objective is to promote open-weight AI that can operate locally on devices, reducing reliance on proprietary AI labs and their demand for more computing power. The 2-billion-parameter model is tailored for vision and language tasks, enabling real-time object recognition in smart glasses. However, no smart glasses equipped with PrismML have been announced as of now.",
  "summary": "Prism's larger goal is open-weight AI that runs on devices and makes better use of the computing power they already have.",
  "key_points": [
    "PrismML develops 1-bit Bonsai LLM for smart glasses",
    "Model reduces larger AI models by 4x while maintaining performance",
    "2-billion-parameter vision and language model for real-time object recognition"
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}