Opera has begun an opt-in WebAI preview in Opera One, letting users run supported AI models locally instead of relying on cloud processing.
Opera introduced WebAI, a browser feature designed to run AI models on a user’s own device rather than send model requests to the cloud. As of October 27, the feature is available as an opt-in early preview in Opera One, the company’s flagship browser.
Opera says WebAI can work with multiple models rather than being tied to one provider. Examples it identifies include Google’s Gemma open-weight model and, on Mac computers, Apple’s on-device foundation model. The company also describes browser-based tasks such as generating text, classifying content, and identifying relationships between pieces of information.
The local approach changes where the processing takes place. If a task is handled on the device, users need not send that information to a cloud provider for that operation, and developers can add AI features without maintaining their own AI infrastructure or paying for each request. Opera also says users can access the features without an AI subscription. These are potential advantages, not a guarantee that every privacy concern disappears: the company notes that local model performance depends on the user’s hardware.
WebAI builds on Opera’s earlier work bringing locally run models into its browsers. The current preview is limited to Opera One, and users must choose to enable it. Opera says its gaming-focused Opera GX and AI-agent browser Opera Neon are expected to receive the feature soon, but it has not given a specific date in the article.
The model options are described as open-weight, which means the model weights are available for use and customization, but that does not necessarily make a model fully open-source or remove all restrictions. Running models locally also uses the computer’s own resources and may affect performance. For developers, the preview offers a way to explore browser-based AI without routing every interaction through a company server; for users, the experience will depend in part on which models and devices are supported.