ToolsJuly 15, 2026via MarkTechPost
Google Releases LiteRT.js: A JavaScript Binding of LiteRT That Runs .tflite Models in Browsers via WebGPU
Why it matters
LiteRT.js enables on-device inference at scale in web applications, reducing server dependency and latency for real-time ML features. This shifts the cost and complexity of model deployment from backend to client-side infrastructure.
Key signals
- Released July 9, 2026
- JavaScript binding of LiteRT for browser execution of .tflite models
- Up to 3x performance gains over competing web runtimes
- 5–60x speedup for GPU/NPU vs CPU inference path
- WebAssembly + XNNPACK (CPU), WebGPU (GPU), experimental WebNN (NPU) support
- Manual tensor memory management required
- Enables on-device ML inference without server calls
The hook
Google just shipped LiteRT.js. Now ML models run natively in browsers at 3–60x faster.
Google released LiteRT.js on July 9, 2026. It is a JavaScript binding of LiteRT, Google's on-device inference library. The runtime executes .tflite models directly in the browser through WebAssembly, with XNNPACK on CPU, ML Drift over WebGPU, and experimental WebNN for NPUs. Google reports up to 3x gains over other web runtimes, and 5–60x for GPU or NPU over its own CPU path. One detail the announcement omits: tensors are manually managed and must be deleted.