Opera is previewing WebAI, a standard for running AI models inside its browser instead of sending requests to cloud services. The approach gives users another way to access browser-based AI while keeping the information processed by the model on their own device.
How WebAI brings AI into the browser
WebAI is designed to work with a range of models rather than being limited to one system. Opera cites Google’s Gemma, an open-weight model, as an example. On Macs, the standard also supports Apple’s on-device foundation model, the company says.
The browser maker says WebAI grew from earlier work to bring locally run models into its browsers. Its early preview is available in Opera One, Opera’s flagship browser. People must opt in to use it.
Opera says its gaming and streaming browser, Opera GX, and Opera Neon, which uses AI agents to perform tasks, are expected to receive the same features soon. For now, the preview gives Opera One users a way to try AI processing that takes place locally.
What users and developers can do
According to Opera, WebAI can generate text, classify content and identify how pieces of information relate. Those capabilities can support AI features within the browser without requiring each request to be handled by a company’s cloud infrastructure.
That arrangement may matter to developers as well as users. Opera says developers can add AI features without operating AI infrastructure or paying for every request. Users, in turn, can use local AI without an AI subscription, according to the company.
WebAI is built to support open-weight AI systems. In these models, the underlying weights are publicly available, which gives developers more room to customize how the models work. They are typically free to download and use, though open-weight models can still have restrictions and are not the same as fully open-source software.
Privacy benefits come with limits
Opera presents local processing as a privacy advantage: information handled by the AI stays on the user’s computer rather than being sent to a company’s servers. That can reduce the amount of personal information a user needs to share with a cloud-based AI service and removes the need for a cloud AI subscription to use these features.
Keeping processing on a device does not erase every privacy risk. It does, however, give users more control over where the AI work takes place. That distinction is relevant as AI assistants gain access to personal information and users may not fully understand how services store or use what they share.
The opt-in setup also leaves the choice with the user: people can decide whether they want to enable the preview. The model options and local processing make WebAI a different route to AI features, while the underlying privacy and usage conditions can still depend on the models involved.
Local AI depends on the computer
Running a model locally shifts the work to the user’s own machine. That can put more demand on the computer and affect its performance. Opera notes that how well models perform will depend on the device’s hardware.
For that reason, the experience may vary from one computer to another. WebAI’s promise is that models can run without relying on a cloud service, but the practical result depends in part on whether a user’s hardware can handle the workload.
Opera’s preview brings together several trade-offs: local processing can offer more control and avoid cloud fees, while using a model places demands on the device. As the feature reaches more of Opera’s browsers, those considerations will shape how useful browser-based AI is for different users.