AIDigest/2026/07/12/2026-07-12-06-google-litert-js-web-ai-inference

LiteRT.js Brings Google's On-Device Inference Engine to the Browser

Source: Google Developers Blog — 2026-07-09

Summary

Google launched LiteRT.js, bringing its on-device LiteRT inference engine to the browser via WebGPU, WebNN, and WASM backends. Google claims up to 3x faster inference than other web AI runtimes, with a 5-60x GPU/NPU speedup over CPU-only execution.

Key Takeaways

Brings LiteRT (Google's on-device inference engine) to the browser, not just native mobile/edge apps.
Supports WebGPU, WebNN, and WASM backends, picking the best available acceleration path per browser/device.
Claims up to 3x faster than other web AI runtimes and 5-60x GPU/NPU speedup versus CPU-only inference.
Lowers the bar for shipping real on-device AI features directly in web apps, without a native app or server round-trip.

Discussion

(No questions yet — ask follow-ups via a Claude Code chat session on this repo; answers get appended here.)