Opera brings on-device AI models from Google and Apple directly into its browser

Opera launched an early preview feature in its Opera One desktop browser that allows users to run artificial intelligence models locally on their machines without cloud processing. The functionality is powered by WebAI, a proposed web standard supported by Google and Microsoft that lets extensions and websites hook directly into hardware-hosted models. At launch, the integration supports Google's Gemma and allows Mac users to tap directly into Apple's on-device foundation model, with rollouts to Opera GX and Opera Neon planned next. [1][2]

The feature is opt-in, free of subscription fees, and gives users granular control over which websites can access local processing. Because models run directly on the device, extensions can handle tasks like organizing tabs or summarizing text without transmitting private personal data to remote servers. However, Opera points out that local execution relies heavily on the user's computer hardware, which can affect overall machine performance. [1][2]

Alongside WebAI, Opera introduced an experimental feature called HTML-in-Canvas, which lets developers place fully functional, interactive web pages into 2D and 3D virtual spaces. The tool enables standard web elements like forms and buttons to remain usable inside real-time graphics scenes, which the company showcased through interactive 3D demonstration environments. [1]