# On-Device — AI updates

Small and on-device AI — models compressed to run on phones, browsers, and the edge, away from the API.

- [Small, on-device AI models gain traction globally](https://www.parallelquant.com/posts/small-on-device-ai-models-gain-traction-globally-818b6d) (2026-07-06, IEEE Spectrum): IEEE Spectrum reports growing global adoption of small, low-power AI models that run without a network connection, citing Nigerian startup founder Adebayo Alonge's RxScanner, a handheld spectrometer that identifies counterfeit medication via AI, as an example of a device shrunk to run entirely on an Android phone after cloud latency made it impractical in the field.
- [Google releases LiteRT.js for running AI models in browsers](https://www.parallelquant.com/posts/google-releases-litert-js-for-running-ai-models-in-browsers-06ef1d) (2026-07-15, MarkTechPost): Google released LiteRT.js on July 9, 2026, a JavaScript binding for its on-device inference library LiteRT that runs .tflite models directly in browsers. It uses WebAssembly with XNNPACK on CPU, ML Drift over WebGPU, and experimental WebNN for NPUs, with Google reporting up to 3x gains over other web runtimes and 5-60x for GPU/NPU versus its own CPU path.

---
Published by Parallel Quant — https://www.parallelquant.com