Opera introduces WebAI, a web standard for running AI models locally in the browser

WebAI is designed to work with a range of AI models rather than a single one. Opera names Google's Gemma, an open-weight model Google introduced two years ago, as one example, and says WebAI also supports Apple's on-device foundation model on Macs. Opera says the standard grew out of its earlier work on locally run models in its browsers.

Users must opt in to the feature, according to the company. Opera says local processing means personal information does not have to be sent to a company's servers, and users do not need a cloud AI subscription. The company says WebAI can generate text, classify content and analyze how pieces of information relate, all within the browser.

Opera notes that performance depends on the device's hardware, since running models locally places more demand on the computer.

Developers can add AI features to a web page using models that run on the user's device, without operating inference infrastructure or paying per request. Support depends on the user's hardware and on the browser, so agents built on WebAI will need a fallback path for other browsers and weaker machines.

Open-weight models carry license restrictions even though their weights are public, so teams should check terms before shipping them.