Opera proposes WebAI, a standard for running AI models inside the browser

Opera has introduced WebAI, a proposed web standard that lets AI models run directly in the browser on the user’s own machine rather than in the cloud. It is not tied to one model: Opera mentions Google’s open-weight Gemma as an example, and on Macs it also supports Apple’s on-device foundation model. Open-weight models publish their trained weights so anyone can run them, often with license limits that open-source software does not carry. WebAI is an opt-in early preview in Opera One for now, with Opera GX and the agent-focused Opera Neon next, and it grows out of Opera’s longer experiments with locally run models.

Opera’s pitch is simple: developers can add AI features without running infrastructure or paying per request, and users get AI without a subscription and without their data leaving the device.

A lot of everyday AI work, such as classifying text, summarizing a page or spotting relationships between pieces of information, does not need a frontier model in a distant data center. Running it locally is cheaper, keeps working when a service changes pricing and keeps personal data on the device. Small, specialized local models are multiplying, including compact decision models from AWS and others (Strands Decider). For people who have been pouring sensitive details into cloud chatbots, often with data terms they did not read (Google’s ad-targeting language), that is a real improvement.

The word that matters is standard. A feature in one browser is a product; a standard matters only if others adopt it, and Opera has a small share of the market. If Chrome, Safari and Firefox ignore WebAI or build incompatible versions, developers will have to write for several or skip local AI altogether.

The trade-offs are real. Local models load the user’s hardware, and Opera says performance depends on the device. People with newer machines get fast, private AI; people on older or cheaper hardware get slow or degraded features, so a web that assumes local AI could become less usable for those who can least afford upgrades. Local also does not mean risk-free: a browser model can still be fed malicious page content, make mistakes or be used by a site in unexpected ways. Keeping data on the device is a privacy gain, not a full security model, and opt-in is the right call. More in our AI coverage and tools coverage.