Opera unveils WebAI standard to run AI models locally in the browser

Opera wants your browser to do its own thinking. The company has introduced WebAI, a new web standard that lets AI models run directly inside the browser instead of pinging a distant data center.

That matters because most AI features today are rented, not owned. Every prompt travels to someone else’s server, and somebody pays the meter. Opera is betting a lot of that work can happen on the machine already sitting on your desk.

What Opera is actually shipping

Opera released WebAI as an early preview on October 27, 2026. The first home for it is Opera One, the company’s flagship browser.

Opera GX and Opera Neon are slated to get the feature later. For now, it targets 64-bit Windows machines and Macs running Apple silicon.

The standard is built to be model-agnostic, meaning it is not tied to a single AI system. Opera points to Google’s Gemma as one example of a compatible model.

Apple’s on-device foundation model is also supported at launch, on compatible Macs. That gives Mac users a path to local AI using a model their operating system already provides.

On the capability side, WebAI covers text generation, content classification, and spotting relationships between data points. Put more plainly, a website or extension can ask the browser to write something, sort something, or figure out how pieces of information connect.

Why developers might care more than users

The pitch to developers is mostly about cost and plumbing. Opera says WebAI lets them add AI features without maintaining cloud infrastructure or paying per-request fees.

Privacy is the other half of the sales pitch. Because processing stays local, the data being analyzed does not have to leave the machine.

Opera is also leaning on permissions. WebAI is disabled by default, and users control activation through granular permissions. Sites do not get to quietly spin up a model on your laptop without asking first.

Opera has floated a few use cases to make this concrete. They include organizing browser tabs, detecting clickbait, and improving how job listings show up inside extensions.

A two-year runway to get here

This is not Opera’s first experiment with local AI. The company began testing on-device large language models in 2024, first showing local LLM support through its developer channel in April 2024.

WebAI is an attempt to turn that tinkering into something more structured: a standard that other developers can build on predictably.

Opera also frames WebAI as a collaborative effort, saying it has backing from Google and Microsoft.

What this means for the browser and AI landscape

For developers, the trade-off is about reach versus cost. Building on WebAI today means targeting a subset of users: those on Opera One, with 64-bit Windows or Apple silicon Macs, who have opted in.

Opera’s mention of Google and Microsoft is the detail to watch here, because their browsers reach far more people than Opera’s do. Whether that backing translates into support beyond Opera’s own products will largely determine whether WebAI becomes infrastructure or stays a niche feature.