Opera introduces a new web standard to let users deploy AI locally in their browser

Introduction

Opera has unveiled WebAI, a web standard that enables AI models to run directly inside its browsers without sending data to the cloud. By leveraging open‑weight models such as Google’s Gemma and Apple’s on‑device foundation model, WebAI aims to give users greater privacy, lower subscription costs, and developers a flexible way to embed AI features.

How WebAI Works

WebAI is built into Opera One as an early preview and will roll out to Opera GX and Opera Neon. The standard provides a set of browser APIs that let developers:

• Load model weights that are publicly available.
• Perform inference locally on the user’s CPU or GPU.
• Access common AI tasks—text generation, classification, and relational reasoning—directly from JavaScript.

Because the computation happens on the device, the user’s data never leaves the browser, eliminating the need for cloud‑based AI subscriptions.

Benefits for Users and Developers

Privacy and Control

Running models locally means personal prompts and results stay on the user’s hardware. This reduces exposure to data‑collection practices common in cloud AI services and aligns with growing privacy concerns.

Cost Efficiency

Open‑weight models are typically free to download, and developers avoid paying per‑request fees for cloud inference. Users also sidestep subscription fees tied to commercial AI platforms.

Flexibility

WebAI is model‑agnostic. While Opera highlights Gemma and Apple’s foundation model as examples, any open‑weight model that conforms to the API can be used, allowing developers to customize behavior or fine‑tune models for niche use cases.

Trade‑offs and Performance Considerations

Local inference demands CPU/GPU resources, so performance will vary by device. High‑end desktops may handle large models smoothly, whereas older laptops or mobile devices could experience slower response times or higher power consumption. Opera notes that developers can choose model sizes appropriate for their audience, and future browser updates may include hardware‑acceleration optimizations.

Looking Ahead

WebAI positions Opera as a privacy‑first browser in an era where AI assistants are increasingly data‑hungry. If the standard gains traction, other browsers may adopt similar capabilities, potentially reshaping how web‑based AI is delivered. For now, early adopters can enable the feature in Opera One’s settings and experiment with open‑weight models, while developers can start integrating the new APIs into their web apps.

--- Opera’s WebAI is an early step toward decentralized AI on the web, offering a compelling alternative to cloud‑centric models for users who value privacy and cost savings.