WebGPU is a web standard that lets web pages use the computer's graphics chip (GPU) for fast graphics and general computing, including running AI models in the browser.
In plain words
Its predecessor, WebGL, was built for drawing and handles general-purpose GPU computing poorly, yet machine learning needs exactly that. WebGPU is designed to match modern native GPU interfaces such as Direct3D 12, Metal and Vulkan, and supports general computation. Libraries such as Transformers.js use it to run models directly in the browser, on the user's own device. Not every browser supports it yet.
A quick example
Imagine a web page that transcribes speech without sending audio to a server. It downloads a small model once, and Transformers.js runs it on the visitor's GPU through WebGPU, which the developer switches on by choosing the webgpu device when loading the model.
Where you'll meet it
- Running AI models in the browser, on-device.
- Complex graphics drawn in web pages.
Related
Sources
- DocsWebGPU API, MDN Web Docs · read 29 Sept 2026
- DocsWebGPU, W3C · read 29 Sept 2026
- DocsRunning models on WebGPU, Hugging Face Transformers.js documentation · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.