Google releases LiteRT.js for running AI models in browsers

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.

Why it matters: Gives web developers a faster path to run AI models client-side without a server.