Opera ships local AI models in its browser via the WebAI standard

Opera has started shipping local AI models directly inside its browser, using the newly proposed WebAI web standard, so that user queries and data stay on the local machine instead of traveling to the cloud. The company confirmed the feature is live as an early preview on Opera One for Windows PC (64-bit, Intel/AMD) and Macs with Apple silicon, with no support yet for Intel-based Macs, Linux, or Windows on Arm. The release removes a barrier that has slowed local AI adoption for years: users no longer need to install or configure tooling such as Ollama to run models inside a browser.

The announcement matters because it inverts the dominant architecture of browser AI. Most browsers today ship AI features that depend on cloud services for responses, which means a user's data is not limited to the local machine and could be used by third-party organizations to train large language models (LLMs, the systems that generate text from prompts) or even to build user profiles. Opera is positioning itself against that trend, betting that small, capable open-weight models (models whose trained parameters are publicly available for anyone to download and modify) and increasingly powerful consumer hardware make data-center-free AI viable for everyday tasks.

What did Opera actually release?

Opera's implementation of WebAI is model-agnostic and built around open-weight models, which means users can select models from providers such as Google and Apple. Because the models are open-weight, Opera says anyone can build on top of them without running expensive AI infrastructure or paying for every request — a claim the company argues will translate into more innovation for the Opera browser ecosystem.

The feature is opt-in. Nothing is downloaded until the user explicitly chooses to do so, which addresses a common concern with bundled AI: unwanted background downloads and silent data processing. Developers who want to experiment with the API can consult example extensions published under the name "WebAI: Local AI for the web, now in Opera."

The rollout has clear hardware boundaries, according to Opera:

• Windows PC, 64-bit, Intel/AMD — supported
• Macs with Apple silicon — supported
• Intel-based Macs — not supported
• Linux — not supported
• Windows on Arm — not supported

Those constraints reflect the compute demands of running generative models locally; Apple silicon's unified memory architecture and modern Intel/AMD systems are the current baseline Opera considers workable for an early preview.

Who is behind the WebAI standard?

WebAI is a newly proposed web standard backed by companies including Google and Microsoft. It allows websites and extensions to use browser-hosted AI — that is, AI models that the browser itself runs on the user's machine, exposed to web pages through a standard interface rather than through each site's own cloud backend.

The stakes are structural. If WebAI gains traction, browsers become the distribution layer for AI models the way they already are for web content. Google and Microsoft, both of which operate major cloud AI businesses, are backing a standard that could shift some AI workloads off their servers and onto user devices. For users, the standardization effort means a website or extension could request AI capabilities without the user's data leaving the device; for developers, it means one API surface across participating browsers rather than a fragmented set of vendor-specific integrations.

Opera's model-agnostic approach is a deliberate differentiator. Rather than locking the implementation to a single vendor's model, the browser lets users choose among open-weight options, including models from Google and Apple.

What does Opera's leadership say?

Krystian Kolondra, EVP Browsers and AI at Opera, framed the release as the culmination of work that began with developers. In a statement, Kolondra 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." He added: "We first shipped local models to developers in 2024. Now we're bringing them to everyone. More models and tools are coming, and you'll always decide what runs on your machine."

The 2024 developer release is a useful anchor for the timeline. Local models in Opera are not a fresh experiment but a graduated program: developer-facing first, consumer-facing second, with additional models and tools promised. Kolondra's closing point — that users retain control over what executes on their hardware — is the operating principle behind the opt-in download model.

How does this compare with Firefox?

The ZDNET reporter covering the release, a self-described five-year Opera default user on Linux, macOS, and Android, notes that the problem so far has been that using local AI with browsers is not always easy. Firefox made it possible through its Smart Window feature, but it is not set up to use local AI by default. Users must choose a custom provider and have Ollama (a popular open-source tool for running models locally) installed and running on their machine.

That friction is precisely what Opera is eliminating. The browser handles model hosting itself, which removes the installation, configuration, and process-management steps that keep most non-technical users away from local AI. The reporter's verdict is blunt: Opera is ready to make local-first browsing a reality with less effort than any current alternative.

How do you try it?

According to the source, enabling local AI in Opera takes three steps:

• Download or update Opera One
• Go to Settings > Opera AI and turn on "Enable Web AI API"
• Install an extension and open it from the extensions menu

Opera stresses that the feature is an early preview, so users should expect limitations in model selection, performance, and platform coverage as the implementation matures.

Why does local AI in a browser matter?

The market context is straightforward. Cloud-based AI carries recurring per-request costs for providers and persistent privacy exposure for users; local AI carries neither, at the price of hardware requirements and smaller model sizes. Open-weight models have become smaller and more capable, as the source notes, while consumer hardware — particularly Apple silicon machines — has grown more powerful. The crossover point where everyday AI tasks run acceptably on a laptop has arrived, and Kolondra's claim that "a lot of what people want from AI doesn't need a data center" is the strategic bet Opera is placing on that crossover.

For the browser market specifically, Opera was one of the first browsers to integrate AI into its interface, making AI features easy and free to use, per the source. Local AI is the next competitive frontier: the vendor that makes on-device AI frictionless gains a privacy story that cloud-dependent rivals cannot match without changing their architecture. Opera has also been shipping adjacent defensive features, such as detection and blocking of malicious clipboard content, and recently released its Air browser — signals of an active development cadence around safety and AI.

What comes next?

Kolondra's statement points to a roadmap: "More models and tools are coming, and you'll always decide what runs on your machine." The immediate watch items are the expansion of supported platforms — Linux, Windows on Arm, and Intel Macs are all absent from the preview — and whether Google and Microsoft's backing of the WebAI standard produces implementations in Chromium and Edge. If WebAI becomes a cross-browser standard, Opera's head start on consumer-facing local AI could set the template the rest of the market follows.