Opera has released an early preview of WebAI, a proposed web standard that lets desktop browsers run open-weight artificial intelligence models entirely on local hardware. The feature launches first in Opera One for Windows and Mac, removing the need for cloud servers and subscriptions.
Desktop and laptop hardware now handle the heavy lifting of artificial intelligence, moving the process away from remote data centers. This setup aims to eliminate cloud latency and per-request fees while keeping private prompts off corporate servers.
WebAI Runs on Local Hardware
The browser’s on-device processing capabilities graduated to an everyday preview after first rolling out in developer channels in 2024. The implementation is model-agnostic, supporting Google’s Gemma and, on compatible Mac computers, Apple’s on-device foundation model.
The feature is disabled by default. Users must manually opt-in through browser settings before any models are fetched or executed on the machine, and granular permission controls let individuals decide which websites gain access to the local intelligence layer.
Eight APIs Enable Developer Tools
The software introduces an interactive playground featuring eight built-in application programming interfaces to show the utility of on-device processing. Developers can use these tools for text and data operations without maintaining backend infrastructure:
- Prompt API for conducting real-time chats governed by custom system rules.
- Summarizer API to shrink long blocks of text into concise headlines.
- Writer API designed to draft new text directly on the fly.
- Rewriter API for making instant tone adjustments.
- Translator operating entirely offline.
- Proofreader for catching and flagging grammar corrections.
- Semantic Embedder converting sentences into mathematical vectors.
- Decisions API for handling quick multiple-choice sorting tasks.
Extensions built on these APIs could organize cluttered browser tabs, automatically detect clickbait, or highlight relevant job listings straight on a webpage.
Opera Launches HTML-In-Canvas for Interactive 3D Environments
Opera is also launching HTML-in-Canvas, which allows developers to drop fully functional, interactive web pages directly into two-dimensional and three-dimensional virtual environments using standard HTML and CSS.
Honestly, this is a fun one. You can take a real web page and throw it into a 3D world, and everything still works: buttons, text, menus, all of it. We wanted users to get their hands on it early and see what they come up with.
Maciej Kocemba, Product Director, Opera GX
Privacy Gains Meet Hardware Demands
Sensitive information does not travel across the internet to a third-party server farm because prompts and data are evaluated locally.
A lot of what people want from AI doesn’t need a data center. Your laptop can already do it, without sending your data anywhere.
Krystian Kolondra, EVP Browsers and AI, Opera
Overall performance depends on the computer’s hardware since execution happens directly on the user’s machine. The preview currently targets 64-bit Windows computers equipped with Intel or AMD processors, alongside Macs running Apple silicon.
Google and Microsoft back the underlying WebAI proposed standard, but it is not yet known if mainstream browsers beyond this initial rollout will make it available.