Opera launches WebAI to run AI models locally in the browser

A new standard for local AI

Browser maker Opera has introduced a new web standard called WebAI. This standard is designed to let users run artificial intelligence models directly within their web browser. By doing so, the software does not have to rely on external cloud computing to function.

The company is launching this technology as an early preview in its flagship Opera One browser. Opera plans to bring these same capabilities to Opera GX, which is tailored for gamers, and Opera Neon, a browser designed to use AI agents for specific tasks.

Users are not automatically enrolled in this new system. According to the company, people must choose to opt into the feature to begin using the local AI capabilities. This approach allows for a gradual rollout as the technology matures.

How WebAI handles models

WebAI is built to be flexible rather than being tied to one specific AI provider. It is designed to support a wide variety of open-weight AI systems. Open-weight models are AI systems where the underlying weights are publicly available, allowing developers more freedom to customize the software.

As an example of supported technology, Opera name-dropped Google’s Gemma. This is an open-weight model that Google first introduced two years ago. Additionally, the standard supports Apple’s on-device foundation model specifically for users on Mac computers.

The standard enables a variety of tasks directly in the browser environment. Users can use it to generate text, classify different types of content, and understand how various pieces of information relate to one another. All of these processes happen on the user's own hardware.

Privacy and developer benefits

One of the primary goals of WebAI is to address growing concerns regarding digital privacy. Many people surrender large amounts of personal information to AI chatbots without knowing how that data is stored. Local models offer an alternative because the processing stays on the user's device.

By keeping the information on the local machine, users do not have to send their personal data to a company's servers. This removes the need for users to pay for a cloud-based AI subscription to access these features. It gives the user more control over the information the AI handles during a session.

The standard also provides significant advantages for software developers. They can add AI features to their own applications without having to run their own AI infrastructure. As the company stated, "Developers can add AI features without running AI infrastructure or paying for every request."

Hardware trade-offs

While running AI locally offers many benefits, it does come with technical trade-offs for the user. Because the computer is doing the heavy lifting, it puts a higher demand on the device's hardware. This can result in different performance levels depending on what the user owns.

Opera notes that the actual performance of the AI will depend on the specific hardware of the device. A computer with more powerful processing capabilities will likely handle the models more smoothly than a less powerful one. Users should consider their current machine's specs when deciding to use these features.

Despite these demands, the move aligns with Opera's long-standing focus on digital privacy. By moving the computation away from the cloud, the company is leaning into its identity as a privacy-focused browser. This provides a path for users who want the utility of AI without the risks of constant data transmission.

The future of local AI

WebAI grew out of previous work by Opera to bring locally run models into the browsing experience. This project represents a step toward a more decentralized way of interacting with artificial intelligence. It sets a foundation for how other web standards might handle local machine learning in the future.

The rollout will continue as Opera integrates the standard into its broader ecosystem of browsers. While the early preview is currently in Opera One, the upcoming additions to Opera GX and Opera Neon will expand the reach of the technology. This will make local AI more accessible to different types of internet users.

As more open-weight models become available, the variety of tasks users can perform locally will likely grow. The standard is built to be compatible with many systems, meaning it can evolve as the AI landscape changes. This ensures that the browser remains a versatile tool for both privacy-conscious individuals and developers.