Opera brings local AI models to its browsers

Opera announced on October 27, 2026 that it is bringing AI models directly into the browser. The new capability, called WebAI, is initially available as an optional early preview in Opera One. Opera GX and Opera Neon are expected to follow. The goal is to make models that run on the user’s own computer, instead of in the cloud, a standard part of the browser.

What WebAI offers in its first release

According to Opera, WebAI is a web standard proposal also backed by Google and Microsoft. The browser works independently of the model and favors open models. Once enabled, Google’s Gemma models can be used directly from Opera, while on Macs Apple’s on-device base model is available. Text generation, content classification, and resolving relationships between pieces of information are among the capabilities in this first release.

No separate app or API key required

WebAI is off by default. No model is downloaded unless the user enables it in Settings, and it can be turned off again after being enabled. Which sites can access local AI is controlled by per-site permissions. Because Opera has built local inference into the browser itself, there is no need to install a separate app, configure a local server, or obtain an API key.

On the developer side, three APIs are offered: Prompt, Decision, and Semantic Embeddings. They are joined by Summarizer, Writer, Rewriter, Translator, and Proofreader APIs. All of them come with in-browser, on-device playgrounds. According to Opera, this setup requires no subscription, keeps processed information on the device, and lets developers add features without running AI infrastructure or paying per request.

A follow-up to the 2024 developer experiment

Opera first opened local models to developers in April 2024. At the time, 150 variants from around 50 model families, including Llama, Vicuna, Gemma, and Mixtral, had been added to the developer version of Opera One. Each variant typically took up between 2 and 10 GB. The company described that release as the first major browser to offer built-in access to local AI models.

Which devices it runs on

For now, the early preview is available on 64-bit Windows PCs with Intel and AMD processors, and on Apple silicon Macs. Intel-based Macs, Linux, and Windows on Arm are not on the list. When WebAI is enabled in Settings, a compatible model is downloaded. Opera says this step may take a few minutes and that performance will vary depending on the hardware.

Opera also published six experimental extensions. Project Lens colors the paragraphs of an article according to how well they fit a project, Got It? generates a short quiz about the text you have read, Questionable Intelligence turns questions about the page into a graph, Semantic Art turns text into animated visuals, RecordIn Alchemy turns text into a social media post, and C·I·A·O sends two ASCII cats off to improvise.

The second innovation announced the same day

Opera also announced HTML-in-Canvas on October 27. This feature allows real, interactive web pages to be placed inside 2D and 3D scenes. Previously, canvas content was only drawn as pixels and could not be clicked, typed into, or selected. With the new approach, menus, buttons, and forms move along with the scene but keep working. Opera positions this for game interfaces and for Opera GX, which is aimed at gamers.

Opera’s move is the browser version of the broader trend of taking AI out of cloud subscriptions and putting it on the device. But the fact that it is an early preview, the lack of Linux support, and the fact that model downloads depend on the hardware show that there is still some way to go before widespread use.

Source: Unite.AI