Opera ships on-device AI models in its browser via new WebAI standard

:

Key Points

1. Opera began running AI models locally inside its browsers through a proposed web standard called WebAI.
2. Google and Microsoft are among the companies backing the standard, which works without cloud servers.
3. Developers avoid per-request AI costs and user data stays on the device, Opera said.

The latest:

AI models now run directly inside Opera’s browser rather than on remote servers, the Norwegian company announced on October 27, 2026, through a proposed web standard called WebAI that Google and Microsoft are among the backers of. The Nasdaq-listed firm, trading as OPRA, said sites and browser extensions can tap models without any cloud dependency. An early preview ships first on Opera One.

Details:

• How it rolls out: Opera said the capability arrives first as an opt-in early preview in Opera One, limited to 64-bit Windows machines running Intel or AMD chips and Macs with Apple silicon. The company said Opera GX and Opera Neon will follow later, without naming a date for either browser.
• Not one model: WebAI is not tied to a single system, according to Opera. At launch it supports Google’s open Gemma model, and Apple’s on-device foundation model on Mac hardware. The company framed the standard as a shared layer rather than a vehicle for one vendor’s AI.
• What users get: The feature lets users generate text, classify content and work out relationships between different pieces of information directly inside the browser, Opera said. Because processing stays on the machine, the company said users do not need AI subscriptions to access those functions.
• The developer pitch: According to Opera, developers can add AI features without standing up their own AI infrastructure and without paying per request. That removes the two recurring costs — hosting and inference billing — that currently sit behind most browser-based AI features.
• The privacy claim: Opera said user data is not sent to the company’s servers because computation remains on the user’s own device. The pitch positions local inference as a privacy argument against cloud processing, the default architecture for most consumer AI tools today.
• The executive line: Krystian Kolondra, Opera’s executive vice president for browsers and AI, said: “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.”
• The track record: Opera was the first major browser to offer built-in access to local AI models, in a developer preview released in 2024. The company has been building toward on-device inference for two years before putting it behind a proposed standard.
• The hardware catch: Local model performance depends on the capability of the user’s own machine, Opera said. That makes the experience uneven by design: the same browser feature will behave differently across laptops, unlike a cloud service where the server sets the floor.
• A second launch: Opera released a separate feature the same day called HTML-in-Canvas, which lets developers place real interactive web pages inside 3D scenes while buttons and forms keep working.

Between the lines:

Opera is a small player proposing infrastructure, not just a product. The leverage comes from the backers: with Google and Microsoft named as supporters, WebAI reads as an attempt to make on-device inference a browser-level primitive rather than an Opera differentiator. The hardware dependency Opera itself acknowledges is the limit — a standard whose output quality varies by laptop is harder to build businesses on than a cloud endpoint.

What’s next

Watch whether WebAI advances beyond a proposal toward formal standardization, whether Google or Microsoft ship support in Chrome or Edge, and when Opera moves the preview out of opt-in into Opera GX and Opera Neon.