Opera Wants AI to Run in Your Browser, Not in Someone’s Cloud

October 28 — The browser maker Opera has launched a new web standard called WebAI that lets users run AI models directly in the browser without relying on cloud services.

Opera says WebAI is designed to work with many AI models, citing Google’s Gemma — an open-weights model first released two years ago — as an example. Opera adds that on Macs, WebAI can call Apple’s on-device foundation models.

Opera stresses that WebAI is not tied to any particular model but is built for a range of open-weights AI systems. The project grew out of Opera’s earlier work bringing locally run large language models into the browser.

The feature is now available as an early preview in Opera One, the company’s main browser. Opera says its browser for gamers and streamers, Opera GX, and Opera Neon, which can use AI agents to carry out tasks, will get the same capability soon. Users must opt in to WebAI to use it.

Open-weights models offer several benefits: they are usually free to download and use, and their underlying weights are public, giving developers more freedom to customise them. Unlike fully open-source software, however, open-weights models may still carry some limits on use and modification.

Opera says privacy is another key advantage: users can use AI features without sending personal information to corporate servers or paying for a cloud AI subscription.

“Developers can add AI features without deploying AI infrastructure or paying per request,” Opera said in its announcement. “Users don’t need to buy an AI subscription; information processed by local AI stays on their own device, and they keep control of the AI on their computer.”

Opera also says that with WebAI, users can generate text, classify content and understand the relationships between pieces of information directly in the browser.

As AI assistants grow more capable and can reach more personal information, privacy concerns have been rising. Users often hand over large amounts of personal data to chatbots without knowing how it is stored and used.

Locally run AI models offer an alternative, because all processing happens on the user’s own device rather than on corporate cloud servers. That does not eliminate every privacy risk, but it does give users more control — which is why Opera has for years made digital privacy a central selling point.

There are trade-offs, though. Because the model uses the computer’s own computing resources, it can increase hardware load and affect performance. Opera notes that how well WebAI performs will depend on the user’s hardware.