Opera bets your laptop can handle the AI jobs chatbots send to the cloud

Opera’s desktop browser can now run an AI model on the computer it is installed on and lend that model to the websites and extensions you allow. The feature is called WebAI and it is arriving as an early preview in Opera One. Once switched on, it downloads an open-weight model such as Google’s Gemma, or on a Mac it uses the foundation model Apple already builds into its operating system, and any approved site can ask it to write, sort or compare text.

The practical change is who pays and where your words go. A browser extension that summarizes articles or tidies your tabs with AI usually ships the page to a cloud service, which means an account, an API key or a monthly plan to cover the bill. With WebAI the text stays on the laptop, the developer pays nothing per request and the user pays nothing at all. “A lot of what people want from AI doesn’t need a data center. Your laptop can already do it, without sending your data anywhere,” said Krystian Kolondra, Opera’s executive vice president for browsers and AI.

Developers get three building blocks, which Opera calls the Prompt, Decision and Semantic Embeddings APIs. In plain terms, they let a page generate text, sort information into categories and measure how closely two pieces of content are related. Opera’s own examples are small on purpose: an extension that flags clickbait, one that spots relevant job postings while you browse, one that colors each paragraph of an article by how much it matters to a project you described.

To show the range, Opera released six experimental extensions and says they are demos, not product features. One writes a short quiz to check you understood the article you just read. Another, Semantic Art, turns a page into moving shapes where similar sentences drift next to each other. A few are plainly jokes, such as a tool that rewrites any article as an “insufferably inspirational” post for a parody network called RecordIn, and two ASCII cats that improvise a scene on whatever topic you type.

The bigger ambition sits in the word standard. WebAI is a proposed web standard that, according to Opera, has backing from Google and Microsoft. Chrome already offers developers built-in AI interfaces of its own, running Google’s Gemini Nano model. Opera’s version is model-agnostic: it can use open models that Opera distributes, compatible models already on the machine or Apple’s on-device model. If browsers settle on one shared interface, a developer could write an AI feature once and let it run on whatever model each visitor’s browser carries, instead of renting cloud compute for every reader.

The limits are real. WebAI is a preview, and it only runs on 64-bit Windows PCs with Intel or AMD chips and on Macs with Apple silicon. Linux, Intel-based Macs and Windows on Arm are left out, and Opera has said nothing about its phone browsers. Local models are far smaller than the ones behind the big cloud chatbots, and Opera itself warns that speed depends on your hardware, so an older laptop will answer more slowly than a data center would. Keeping the text on your machine also protects only what the model processes: the website you grant access to still sees what you do on it, as it always has.

Opera has left the switch with the user. Nothing is downloaded until WebAI is turned on in the browser’s AI settings, the first model download can take a few minutes, each website needs its own permission and the whole feature can be turned off at any time. For a company that has sold its browsers on privacy for years and says hundreds of millions of people use them, that opt-in design is the point of the pitch.

The launch, announced from Oslo on October 27, builds on an experimental developer build from 2024 in which Opera became the first major browser to offer built-in access to local models. It arrived alongside HTML-in-Canvas, a separate developer feature that places live, clickable web pages inside 2D and 3D scenes and is aimed at games and Opera GX. WebAI comes to Opera GX, the gamer browser, and Opera Neon, the company’s agent-driven browser, next, with no date given. Opera also plans more open models tuned for different devices and more tools for developers.

Until other browsers ship the same interface, though, a feature built on WebAI works only for people who already use Opera. Whether that changes depends less on Opera than on whether Google and Microsoft carry the proposal from support on paper into their own browsers.

Tags: Apple, Google, Local AI, Gemma, Opera One, WebAI