Browser manufacturer Opera has introduced the new web standard WebAI, designed to allow users to run AI models directly in the browser without relying on cloud services. This standard is compatible with various AI models. Opera demonstrated this with Google’s open-source model Gemma and said that WebAI also supports basic models on Apple devices. This technology does not depend on specific models and is primarily aimed at various open-weight AI systems. It is a continuation of Opera’s previous efforts to explore running large language models locally in browsers.
Currently, WebAI has been launched in the Opera flagship browser, Opera One, and users need to actively enable it. Opera GX for gamers and streamers, as well as Opera Neon that supports AI agents in performing tasks, will also receive these features later.
WebAI supports generating text, classification content, and analyzing relationships between information in a browser. Opera says that local inference can reduce reliance on cloud-based AI infrastructure and paid subscriptions, while allowing user data to remain on personal devices and enhancing privacy control. However, open-weight models are not equivalent to full open source, and local operation cannot completely eliminate privacy risks; actual performance also depends on the device’s hardware configuration.
As the personal information processed by AI assistants continues to increase, on-device AI is becoming an important alternative to cloud-based services. WebAI attempts to integrate model running capabilities more deeply into browsers, but balancing privacy protection, device performance, and model capabilities remains a challenge for the widespread adoption of on-device AI applications.