Opera Launches WebAI So Websites Can Use AI Models Running on Your Own Device
Opera announced WebAI on 27 October 2026, a way for websites and browser extensions to use AI models that run on the visitor's own device rather than in the cloud. It is available now as an early preview in Opera One, with Opera GX and Opera Neon also named in the announcement.
WebAI introduces three APIs. The Prompt API generates text, the Decision API classifies information, and the Semantic Embeddings API finds relationships between pieces of content. Opera's implementation uses open-weight models running on its own inference engine, and the company describes WebAI as a proposed web standard backed by companies including Google and Microsoft. It is not yet an agreed standard, and Opera's own press material notes that the underlying Built-in AI APIs began as a Chrome proposal.
Control sits with the user. WebAI is opt-in, nothing is downloaded until you switch it on under Settings, then Opera AI, and site-by-site permissions decide which pages can reach the local model. Turning it on downloads a compatible model, which can take a few minutes. At launch it supports Windows PCs with 64-bit Intel or AMD processors and Macs with Apple silicon, and it is not available on Intel Macs, Linux or Windows on Arm.
The appeal for web builders is privacy and cost. Content processed on the device does not need to be sent to a cloud AI provider, and a site pays nothing per request. That suits small, bounded jobs such as classifying a support message, suggesting tags, or searching a page by meaning instead of keywords.
The catch is reach. A feature that only works in one browser, on some hardware, after the visitor has switched it on, cannot be something a site depends on. Treat it as progressive enhancement. Build the page so it works without the model, and let the local AI make it better where it is available.
We will be watching whether other browsers adopt the same interfaces, because that is what turns an Opera feature into a web platform feature.