The short answer
Opera has shipped the first browser implementation of WebAI, a proposed standard for running AI models locally inside the browser. Since October 27, 2026, sites opened in Opera One can ask permission to use an on-device model such as Google’s Gemma or, on Macs, Apple’s foundation model to generate text, classify content and relate pieces of information. For teams doing web application development, it is an early signal that the browser itself may become an AI runtime.
What did Opera actually launch?
WebAI gives websites and browser extensions a way to reach AI models that the browser hosts on the user’s device. Opera’s announcement lists three jobs it can do today: generate text, classify content and work out how pieces of information relate to one another. TechCrunch notes that Opera has been experimenting with locally run models since April 2024; the difference now is that web pages, not just the browser’s own assistant, can use them.
The model layer is meant to be swappable. Opera calls its implementation model-agnostic and built around open models, starting with Gemma and, on Macs, Apple’s on-device model, which 9to5Mac highlights as the first time Mac users can call Apple’s model from inside a browser. Opera says more models tuned for specific devices will follow.
How do privacy and permissions work?
The design follows the pattern of camera and microphone access. WebAI is off by default and the browser downloads no model until the user turns the feature on. After that, each website has to request permission to use the local model, and users can control which sites have access. Opera says data processed this way stays on the device. “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 EVP of Browsers and AI.
Hardware is the practical limit. ZDNET reports the preview runs only on 64-bit Windows PCs with Intel or AMD chips and on Apple-silicon Macs, and how fast a model answers depends on the visitor’s machine. A feature that responds in a second on a new MacBook may be slow or unavailable on an older office laptop.
Why does a browser AI standard matter to developers?
It changes who pays for inference and where data goes. Today most AI features in web products call a hosted model through an API, so every summary or classification adds to a per-token bill and sends user content to a third party. Opera’s own pitch is that developers can add AI features “without running AI infrastructure or paying for every request.” For high-volume, low-stakes tasks such as tagging, form assistance or on-page search, local inference could remove most of that cost.
The catch is fragmentation. WebAI is still a proposal, and so far it ships in one browser with a small share of the market. Chrome and Edge have been testing their own built-in AI APIs, and until vendors agree on one interface, teams risk writing features against an API that changes or never reaches most of their users. Output quality will also vary by model and device, which makes testing harder than with a single hosted model.
What it means for US & EU software teams
For EU teams, on-device processing is attractive because personal data that never leaves the browser does not need to be sent to a model provider or covered by a cross-border transfer. It does not remove GDPR duties: you still decide what the feature does with personal data, what you log and how users are informed, so a GDPR compliance review of the data flow is still worth doing before launch.
For US teams, the near-term story is cost and resilience. A local path can take load off paid APIs for simple tasks and keep basic AI features working when a provider is slow or rate-limited. But with support limited to one browser, the business case today is experimentation, not migration.
What should teams do now?
- Map candidate features. List AI tasks in your web app that are high-volume and low-risk: classification, short drafts, semantic search, content tagging.
- Build behind an interface. Put model calls behind one client-side abstraction so a local model, a browser API or a server endpoint can be swapped without rewriting features.
- Use feature detection. Check whether local AI is available and permitted, and fall back to your server model silently when it is not.
- Test on real hardware. Measure latency and quality on the laptops your users actually own, not only on developer machines.
- Keep sensitive logic server-side. Pricing, eligibility, security and anything that must give the same answer for every user should not depend on a model that varies by device.
- Watch the standard. Track whether Chrome, Edge and Safari adopt WebAI before committing roadmap items to it.
None of this is legal advice. The broader direction is clear, though: browsers are starting to treat AI models like cameras or storage, a device capability that sites can request. Teams that design for a hybrid of local and cloud models now will be able to adopt it cheaply if the standard spreads.
Frequently asked questions
What is Opera WebAI?
WebAI is Opera's implementation of a proposed web standard that lets websites and browser extensions use AI models running locally on the user's device through the browser. Opera released it as an opt-in early preview in Opera One on October 27, 2026, and says the proposal is backed by companies including Google and Microsoft.
Which AI models does WebAI run?
At launch Opera supports Google's open-weight Gemma model and, on Macs with Apple silicon, Apple's on-device foundation model. Opera describes the implementation as model-agnostic and built around open models, with more device-optimized models planned.
Which devices and browsers support WebAI today?
Only Opera One, as an early preview on 64-bit Windows PCs with Intel or AMD processors and on Macs with Apple silicon. ZDNET reports it is not available on Intel Macs, Linux or Windows on Arm. Opera GX and Opera Neon are due to follow, and other browsers have not shipped WebAI yet.
Does website data leave the device with WebAI?
According to Opera, information processed by the local model stays on the user's device. The feature is opt-in, nothing downloads until the user turns it on, and each website must ask for permission before it can use the local model, similar to a webcam prompt.
Should we build product features on WebAI now?
Treat it as progressive enhancement, not a dependency. Detect whether local AI is available, use it for low-risk tasks such as classification, drafting or on-page search, and keep a server-side model as the fallback until the standard is adopted beyond Opera.
Sources
Opera Newsroom — Opera doubles down on local AI, shipping models directly in its browsers (primary source, October 27, 2026)
TechCrunch — Opera introduces a new web standard to let users deploy AI locally in their browser (October 27, 2026)
ZDNET (via Yahoo Tech) — Opera will run AI locally in your browser now (October 28, 2026)
9to5Mac — Opera One now lets Mac users run Apple’s AI model directly in the browser (October 27, 2026)