Environment:Tensorflow Tfjs Browser Runtime

Metadata
Source	TensorFlow.js
Domains	Infrastructure, Browser_ML
Date	2026-02-10 00:00 GMT

Overview

Browser runtime environment for TensorFlow.js with WebGL, WebGPU, WASM, and CPU backend support.

Description

Browser-based execution of TensorFlow.js models. Supports four backends:

WebGL (priority 2, auto-registered when isBrowser() is true) — GPU-accelerated backend using WebGL 2.0 or 1.0 shaders. This is the default backend in most browsers.
WebGPU (priority 3, requires Chrome M113+ with navigator.gpu) — Next-generation GPU backend offering improved performance over WebGL. Highest priority when available.
WASM (requires WebAssembly support, optional SIMD and multi-threading) — CPU-based backend compiled from C++ to WebAssembly, significantly faster than pure JS for CPU workloads.
CPU (pure JavaScript fallback) — Reference backend with no hardware requirements. Slowest, but guaranteed to work in any JS environment.

TensorFlow.js automatically selects the highest-priority available backend at startup. The backend priority determines selection order: WebGPU (3) > WebGL (2) > WASM/CPU (1).

Usage

Use this environment for any browser-based ML inference or training workflow. Required by all implementations that use browser APIs like tf.browser.fromPixels(), model.predict(), or tf.loadGraphModel() with HTTP URLs.

System Requirements

Requirement	Details
Browser	Modern web browser: Chrome 79+, Firefox, Safari, Edge
GPU	Required for WebGL and WebGPU backends. Integrated or discrete GPU with up-to-date drivers.
WebGPU	Chrome M113+ (default since May 2023) with `navigator.gpu` available
WASM	Browser must support WebAssembly. SIMD and multi-threading detection is automatic.
CPU Backend	No special hardware requirements (pure JavaScript)

Dependencies

Union Package (Recommended)

npm install @tensorflow/tfjs

The @tensorflow/tfjs union package bundles all browser-relevant packages together.

Individual Packages

Package	Purpose
`@tensorflow/tfjs-core`	Core tensor operations, engine, and backend interface
`@tensorflow/tfjs-layers`	Keras-compatible layers API for model building
`@tensorflow/tfjs-converter`	Loading and executing converted TensorFlow models
`@tensorflow/tfjs-backend-webgl`	WebGL backend (GPU acceleration)
`@tensorflow/tfjs-backend-webgpu`	WebGPU backend (next-gen GPU acceleration)
`@tensorflow/tfjs-backend-wasm`	WebAssembly backend (optimized CPU)
`@tensorflow/tfjs-backend-cpu`	Pure JavaScript CPU backend (fallback)

At least one backend package must be installed alongside the core packages.

TypeScript

TypeScript 5.0.4

Credentials

None required for browser runtime.

Quick Install

Via npm

npm install @tensorflow/tfjs

Via CDN Script Tag

<script src="https://cdn.jsdelivr.net/npm/@tensorflow/tfjs"></script>

Code Evidence

WebGL Backend Auto-Registration (Browser Only)

From tfjs-backend-webgl/src/base.ts:24-26:

if (device_util.isBrowser()) {
  registerBackend('webgl', () => new MathBackendWebGL(), 2 /* priority */);
}

The WebGL backend only registers itself when running in a browser context. The priority of 2 means it is selected by default unless a higher-priority backend (WebGPU at priority 3) is available.

WebGPU Requires navigator.gpu

From tfjs-backend-webgpu/src/webgpu_util.ts:157-159:

function isWebGPUSupported(): boolean {
  return (typeof globalThis !== 'undefined') &&
         (globalThis.navigator.gpu != null);
}

WebGPU support is detected by checking for the presence of navigator.gpu on the global object.

isBrowser Detection

From tfjs-core/src/device_util.ts:60-64:

function isBrowser(): boolean {
  return (typeof window !== 'undefined' && window.document != null) ||
         (typeof WorkerGlobalScope !== 'undefined');
}

The isBrowser() utility returns true when running in a window context or inside a Web Worker.

Common Errors

Error	Cause	Solution
WebGPU not supported	Browser does not expose `navigator.gpu`	Upgrade to Chrome M113+ or use a WebGPU-enabled browser. Falls back to WebGL automatically.
WebGL context lost	GPU memory exhaustion or driver instability	Reduce model size, decrease texture dimensions, or use the WASM backend as an alternative.
EmscriptenModule not found (Angular)	Angular build does not include Emscripten types	Add `@types/emscripten` to the `types` array in `tsconfig.json`.

Compatibility Notes

WebGPU became the default GPU backend in Chrome since M113 (May 2023). It is not yet supported in iOS Safari.
WASM SIMD and threading feature detection is automatic — the backend selects the optimal binary at runtime.
WebGL backend checks isBrowser() before auto-registering, so it will not register in Node.js environments.
iOS Safari supports WebGL only (no WebGPU yet). Performance may be lower than desktop browsers.
Web Workers are supported — isBrowser() returns true inside WorkerGlobalScope.

Related Pages

Required By (Implementations)

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment