Opera previews WebAI standard for local browser AI

Opera brings local AI processing to Opera One

Opera has introduced WebAI, a proposed web standard designed to let people run AI models directly in the browser without depending on cloud infrastructure. The company is making the capability available as an early preview in its flagship Opera One browser, with opt-in required from users.

WebAI is intended to work with a range of AI models rather than one proprietary system. Opera cited Google’s open-weight Gemma as an example, and said Macs also support Apple’s on-device foundation model. The browser maker says the standard is not tied to a particular model and is designed for a broad set of open-weight systems.

The announcement extends Opera’s prior work on bringing locally run models into its browsers. Opera said the same treatment will soon reach Opera GX, its browser for gamers and streamers, and Opera Neon, which uses AI agents to perform tasks.

What WebAI is designed to do

Opera says WebAI can generate text, classify content and help users understand relationships between pieces of information, all within the browser. In this model, the AI processing takes place on the user’s device instead of a company cloud server.

That architecture is central to Opera’s privacy argument. The company says information handled by local AI remains on the device, allowing users to use AI features without sending personal information to a provider’s servers or paying for a cloud AI subscription. Opera also says developers could add AI features without operating AI infrastructure or paying for each request.

Open-weight models can be downloaded and used without charge, while their publicly available model weights give developers more latitude to customise behaviour. Opera notes, however, that open-weight models are not equivalent to fully open-source software and may still carry restrictions.

Privacy benefit comes with a hardware trade-off

Local execution does not remove every privacy risk, but it gives users more control over where processing occurs. That distinction is increasingly relevant as AI assistants handle more personal information and users may not always understand how cloud-hosted chatbot data is stored or used.

There is a practical constraint: running models locally increases demand on the computer. Opera says model performance will depend on the device’s hardware, and local AI workloads can affect performance. The business implication is that teams considering browser-based AI should weigh data-handling requirements and subscription costs against the hardware capacity of the devices that will run the models.