Opera Launches WebAI: Free On-Device AI in the Browser. Does It Handle Arabic?

On 27 October 2026, Opera began shipping local artificial intelligence models directly inside its desktop browsers. The company introduced an early preview called WebAI, which is available today in Opera One, with rollout to Opera GX and Opera Neon coming soon. The release follows Opera’s experimental developer build from April 2024, shifting on-device model execution from a specialized testing environment into mainstream browsing.

Local Inference Without Cloud Servers or API Keys

WebAI enables websites and browser extensions to run AI tasks entirely on a user’s machine instead of delegating computations to remote servers. As reported on the Opera blog, Opera describes WebAI as a proposed web standard backed by companies such as Google and Microsoft. The implementation runs without standalone software, local server setups, or external API keys. At launch, supported options include Google’s open-weight Gemma model alongside Apple’s on-device foundation model on macOS, with promises from Opera to integrate additional open-weight models over time.

Developers access these capabilities through standard interfaces: Prompt, Decision, and Semantic Embeddings APIs for text generation, classification, and relational analysis. Opera also lists Summarizer, Writer, Rewriter, Translator, and Proofreader APIs on its next.opera.com/web-ai playground. Alongside WebAI, Opera introduced HTML-in-Canvas, an unrelated feature that embeds interactive web pages inside 2D and 3D scenes.

Costs, Privacy Guarantees, and User Control

WebAI is completely free to use and requires no paid AI subscriptions. Because processing happens on your own computer, the page content the model handles does not need to be sent to a cloud AI provider. As Krystian Kolondra, Opera’s EVP of Browsers and AI, noted: “A lot of what people want from AI doesn’t need a data center.”

The feature operates strictly on an opt-in basis; no models are fetched until a user explicitly turns the feature on. Furthermore, site-specific permissions ensure users dictate which individual websites can interact with the on-device engine. To illustrate potential applications, Opera shared six experimental extensions, including one converting articles into kinetic art, alongside concepts for tab sorting, job-post filtering, and clickbait detection.

How to Enable WebAI in Opera One

1. Download or update to the latest build of Opera One.
2. Open Settings, navigate to Opera AI, and enable the toggle for Enable Web AI API. This triggers the download of a compatible model, which may take several minutes.
3. Install one of Opera’s example extensions, or open the WebAI playgrounds, where each card shows whether that API works in your browser.

Hardware Limits and Early Preview Caveats

Because WebAI runs on local silicon, performance varies significantly based on device specifications. Opera limits the preview to 64-bit Windows machines with Intel or AMD processors and Apple-silicon Macs. Linux, Intel-based Macs, and Windows on Arm systems are not supported, and mobile platforms are unmentioned. Opera published no details regarding model parameter size, RAM requirements, or storage footprint. As reported by TechCrunch, local models increase system load, which can noticeably impact overall computer performance.

Testing Arabic Dialects and On-Device Data Privacy

Opera mentions no regional limits for WebAI, but it also published no official language list or verification of right-to-left script handling. For Arabic readers, testing the technology directly is essential. Users can open the playgrounds at next.opera.com/web-ai to experiment with Arabic prompts across colloquial dialects like Egyptian or Gulf, evaluate two-way Arabic and English translation, and verify whether right-to-left text renders properly.

The clear advantage here is confidentiality. Working on local models ensures sensitive Arabic drafts, financial summaries, or private documents remain confined to the user’s hard drive without exposure to third-party servers.

Early Verdict and Standalone Alternatives

WebAI is the most practical attempt yet to make private, on-device AI something you switch on rather than install, and it lowers the cost for small developers to zero per request. Still, it remains an early preview whose responsiveness depends entirely on local hardware. For users who prefer running local models outside the browser environment, desktop apps such as Ollama or LM Studio give you more control over which model you run; our Ollama vs LM Studio comparison covers which suits whom. Our advice: if you already use Opera on a supported computer, switch WebAI on and run the Arabic test above before you trust it with real work.