Opera proposes WebAI standard for running AI models locally in the browser

What Opera announced

Opera has introduced a proposed web standard, WebAI, that runs AI models locally inside the browser instead of routing requests through the cloud. According to TechCrunch, the goal is to let sites and developers build AI-powered features that execute entirely on a user's own machine.

The standard is not tied to a single model. Opera cites Google's Gemma, an open-weight model, as one example, and says that on Macs WebAI can also work with Apple's on-device foundation model. More broadly, the company describes WebAI as built to support a wide range of open-weight AI systems. The project grew out of Opera's earlier work on bringing locally run models into its browsers.

Rolling out as an opt-in preview

The new capabilities are shipping as an early preview in Opera One, the company's flagship browser. TechCrunch reports that the same features are slated to arrive soon in Opera GX, the gaming-focused browser, and Opera Neon, Opera's agent-driven browser. Participation is opt-in: users have to actively turn the feature on rather than being enrolled by default.

What WebAI can do

According to Opera, the standard supports generating text, classifying content, and working out how different pieces of information relate to one another, all directly in the browser.

The company frames the economics as a key draw for developers. "Developers can add AI features without running AI infrastructure or paying for every request," Opera said in its announcement, adding that "users don't need AI subscriptions, the information the local AI handles stays on their own device, and they stay in control of the AI on their computer."

The open-weight angle

WebAI's model-agnostic design leans on open-weight AI systems. As TechCrunch notes, these models are generally free to download and use, and because their underlying weights are publicly available, developers have more freedom to customize how they behave. They are not the same thing as fully open-source software, however, and typically come with some restrictions.

Privacy and its limits

The privacy argument is straightforward: when inference happens on the user's device, personal data does not need to reach a company's servers, and users can access AI features without paying for a cloud subscription. TechCrunch points out that this matters at a moment when AI assistants increasingly handle sensitive personal information, often without users clearly understanding how that data is stored or used.

Local processing is not a cure-all. It reduces rather than eliminates privacy risk, but it does shift control back toward the user, which fits Opera's long-standing emphasis on privacy as a selling point.

The hardware trade-off

Local inference carries a cost. TechCrunch reports that running models on-device increases the demand on a user's computer and can affect performance, and Opera itself acknowledges that model performance will depend on the device's hardware. Machines with weaker specifications may struggle to keep up with larger models.

Why it matters

WebAI points at a meaningful shift in how AI could be delivered on the web. If local, in-browser inference becomes a standard capability, developers could ship AI features without provisioning servers or absorbing per-request costs, and users could get those features without subscriptions or sending their data to third parties. That would lower the barrier for smaller sites to add AI functionality and change the economics that currently favor large platforms running cloud-hosted models.

The open question is adoption. Opera has introduced WebAI as a standard, but a web standard gains its power from being implemented across browsers, and TechCrunch's report does not indicate whether other browser makers plan to support it. For now, WebAI is an opt-in preview inside one company's products: a credible technical demonstration of on-device AI on the web, but not yet a shared platform.