Opera Ships Local AI Models in WebAI Preview

Opera has started shipping AI models that run inside its browser on the user’s own computer, rather than on a remote server. The Norwegian company said on 27 October that the first features are available “today as an early preview in Opera One, with Opera GX and Opera Neon to follow”.

The feature is called WebAI. Opera describes it as “a proposed web standard, backed by companies such as Google and Microsoft, that lets websites and browser extensions use AI hosted through the browser”. In practice, a website could ask the browser to summarise, sort or compare text, and the work would be done by a model on the laptop.

“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,” said Krystian Kolondra, Opera’s executive vice president for browsers and AI.

What WebAI gives websites and extensions

According to Opera’s developer blog, “WebAI introduces the Prompt, Decision and Semantic Embeddings APIs”. The press release puts the same idea in plainer terms: the models can “generate text, classify content and understand how pieces of information relate”.

Opera says its version is “model-agnostic and built around open models”. It names Google’s Gemma and, on Macs, Apple’s on-device foundation model. The blog says “there’s no separate AI app to install, no local server to configure and no API key to get”.

To show what is possible, Opera has published six experimental extensions, including one that turns an article into moving art. It calls them “examples to spark ideas, not Opera features”.

Opt-in, and only on some machines

WebAI is switched off by default. “Nothing is downloaded until the user turns it on,” the release says, and users can grant or refuse access site by site. The blog says the setting sits under Opera AI in Settings, where turning it on downloads a compatible local model, and it can be turned off again at any time.

The preview runs on 64-bit Windows PCs with Intel or AMD chips and on Macs with Apple silicon. The blog adds that “it isn’t available on Intel-based Macs, Linux or Windows on Arm”. Opera warns that speed “depends on their hardware” and that “local AI performance is still improving”.

The two Opera texts differ on timing. The blog says the preview is live in Opera One, Opera GX and Opera Neon; the release says GX and Neon are still to follow.

Opera also announced a second experiment, HTML-in-Canvas, which places working web pages inside 2D and 3D scenes. “You can take a real web page and throw it into a 3D world, and everything still works,” said Maciej Kocemba, product director for Opera GX.

Why running the model in the browser changes the cost of AI features

The interesting part of WebAI is who pays. When a website adds a chatbot or a summary button today, each request usually goes to a cloud model, and the site owner pays for it. Opera’s pitch is that “developers can add AI features without running AI infrastructure or paying for every request”, because the user’s own processor does the work.

That moves the cost onto hardware people already own, and it keeps the text on the device, which is the privacy case Kolondra makes. It fits a wider shift that TSN has tracked, from hospital researchers running an open-source assistant on a Mac mini to start-ups releasing coding models built to run locally.

The trade-off is capability. Small local models are slower and less capable than the largest cloud systems, and Opera’s own hardware list leaves out Intel Macs, Linux and Arm-based Windows laptops. A site that relies on WebAI would still need a fallback for everyone else.

A proposed standard needs more than one browser

The bigger question is whether WebAI becomes something web developers can count on. Opera calls it a proposal, not an agreed standard, and the support it cites from Google and Microsoft comes from Opera’s own description. Opera is a small player next to Chrome, which already shapes how the web handles security through features such as its Enhanced Protection warnings. Developers are unlikely to build on WebAI widely unless the larger browsers ship compatible APIs. Until then, Opera’s preview is a working test of whether people will switch local AI on when it is offered as a choice.