Opera has introduced WebAI, a proposed web standard that allows websites and browser extensions to use artificial intelligence models running directly on a user’s computer. Announced on October 27, 2026, the initiative is designed to make local AI more accessible to developers and everyday users without requiring every request to be processed by a cloud-based AI service.
The technology is initially available as an opt-in early preview in Opera One. It supports compatible local models, including Google’s Gemma and Apple’s on-device foundation model on supported Macs. Opera says WebAI can power features such as text generation, content classification and semantic search while keeping the information processed by local models on the user’s device. The company plans to expand availability across its browser lineup.
Key Takeaways
- New standard: Opera has introduced WebAI, a proposed web standard for accessing AI models locally through browsers.
- Local processing: Compatible AI models run on the user’s computer rather than relying on a remote AI service for each request.
- Developer APIs: WebAI introduces APIs for prompting models, making decisions and generating semantic embeddings.
- Privacy benefits: Websites and extensions can process supported content locally without sending it to a cloud AI provider.
- Lower operating costs: Developers can add certain AI features without maintaining their own AI infrastructure or paying for every model request.
- Early preview: The first release is available in Opera One on supported Windows PCs and Apple-silicon Macs.
- Important limitation: Local AI performance depends on hardware, and the technology does not guarantee that every browser activity or website interaction remains private.
What Is Opera WebAI?
WebAI is a proposed web standard that lets websites and browser extensions access compatible AI models through the browser itself.
Most AI-powered web applications rely on a remote service. When a user asks an AI assistant to summarize a document or classify a piece of text, the application typically sends the relevant information to a server, where the model processes it and returns a response.
WebAI offers an alternative for supported tasks. Instead of routing every request to a cloud model, a website or extension can use a compatible model running locally on the user’s device.
Opera says its implementation is model-agnostic, meaning it is not designed around a single AI provider. At launch, users can access Google’s Gemma models and, on supported Macs, Apple’s on-device foundation model.
The initiative builds on Opera’s earlier experiments with local language models in its developer browser in 2024. The new release moves that work toward everyday browser use and gives developers a way to build extensions that use local AI without requiring users to install a separate AI application or configure a local server.
WebAI is still a proposed standard, however. Opera’s implementation and early-preview availability do not mean that every browser or website already supports it.
How WebAI Works
WebAI provides a bridge between websites, browser extensions and AI models available on a user’s computer.
The developer builds a website feature or extension that requests a supported AI operation. The browser manages the local model and performs the inference on the device, subject to the user’s settings and permissions.
Opera has introduced several API capabilities:
| API capability | What it does | Example use |
|---|---|---|
| Prompt API | Sends instructions to a compatible local language model | Summarize an article or generate text |
| Decision API | Evaluates information and returns a structured decision | Classify a message or select a category |
| Semantic Embeddings API | Converts text into numerical representations of meaning | Find related articles or search saved content by topic |
These capabilities allow developers to build more than conventional chat interfaces. An extension could highlight paragraphs relevant to a project, identify job listings that match a user’s interests or help organize browser tabs.
The browser manages access to the local model, so developers do not necessarily need to deploy and operate their own AI inference servers for these features.
The available capabilities depend on the browser version, model and device. Developers must also account for cases where a model is unavailable or a user has not enabled local AI.
Why Running AI Locally Matters
The move toward local AI addresses three important concerns: privacy, cost and control.
Privacy: If a task is processed entirely by a local model, the text or image used for inference does not need to be transmitted to a cloud AI provider. This can be useful when users are working with personal notes, private documents or sensitive browsing information.
Cost: Cloud AI services commonly charge based on usage, model size or the number of requests. With local inference, developers can avoid those per-request model charges for supported operations, although they still need to account for development, testing and other infrastructure costs.
User control: Opera’s implementation is opt-in. Users must enable the feature before compatible models are downloaded, can turn it off later and can control which websites are permitted to access local AI.
These advantages do not mean every aspect of a website becomes private. A website may still send information to its own servers for unrelated purposes, and an extension could have other permissions. Local inference protects the data processed by that local operation from needing to go to an AI provider; it does not replace careful review of website and extension permissions.
Which AI Models Does WebAI Support?
Opera says WebAI is designed to support different compatible models rather than locking developers into one provider.
At launch, the browser supports models from Google and Apple. Google’s Gemma is an open-weight model family that developers can adapt and run in compatible environments. On Macs with Apple silicon, Opera also supports Apple’s on-device foundation model.
This approach gives developers flexibility to select models according to their requirements. A smaller model may be sufficient for classifying text or summarizing a short article, while a more capable model may be better suited to tasks that require stronger reasoning.
The choice also affects performance. Local models use the computer’s own processing resources, including memory and available computing capacity. A powerful computer may run a supported model more comfortably than a lower-specification machine.
Opera says it plans to add more models optimized for different devices, but the number of supported models and their availability may change as the preview develops.
How Developers Could Use WebAI
WebAI could make it easier for developers to add practical AI features to websites and extensions without building a complete cloud AI backend.
For example, a research extension could analyze an article and highlight passages relevant to a user’s project. A browser tool could classify saved pages by topic or identify related documents based on meaning rather than exact keywords.
An extension for job seekers could evaluate listings against a user’s stated preferences. A reading tool could generate a quiz from an article or summarize lengthy content before the user decides whether to read it in full.
Opera has released experimental extensions to demonstrate these possibilities. They are examples of what developers can build with WebAI, rather than all being standard features built into the browser.
For independent developers and small teams, the potential benefit is lower infrastructure complexity. A local AI feature can operate without requiring the developer to pay for every inference request. However, developers still need to design good user experiences, handle errors and ensure that the model’s output is suitable for the task.
WebAI Availability and System Requirements
Opera began rolling out WebAI as an opt-in early preview on October 27, 2026.
| Item | Current details |
|---|---|
| Initial browser | Opera One |
| Windows support | 64-bit Windows PCs using Intel or AMD processors |
| Mac support | Macs with Apple silicon |
| Intel-based Macs | Not supported in the initial preview |
| Linux | Not supported in the initial preview |
| Windows on Arm | Not supported in the initial preview |
| Activation | Enable WebAI in Opera’s AI settings |
| Other Opera browsers | Opera GX and Opera Neon are expected to follow |
Users can try the feature by updating or installing Opera One, opening Settings and navigating to Opera AI, then enabling the Web AI API. The browser downloads a compatible local model, which may take several minutes.
Users can then experiment with compatible extensions or developer demonstrations. Because the model runs on the device, response times depend on the computer’s hardware and the complexity of the requested task.
Developers can find additional information and interactive demonstrations on Opera’s WebAI playground.
The Bigger Picture
Opera’s WebAI initiative reflects a growing effort to move some AI workloads away from centralized cloud infrastructure and closer to the people using them. Smaller, more capable models make it increasingly practical to perform tasks such as summarization, classification and semantic search on consumer hardware.
The browser is a particularly important place for this shift because it already has access to the pages, documents and workflows users want AI to help with. A standard interface for local models could make it easier for developers to add useful features without creating separate AI services for every extension or website.
The challenge will be consistency. Developers need predictable APIs, reliable model performance and clear rules governing how websites can access local AI. Wider adoption will also depend on support from other browsers and the development community. If WebAI gains broader support, it could help make local inference a normal part of the web platform rather than a feature limited to individual applications.
FAQs
1. What is Opera WebAI?
WebAI is a proposed web standard that allows websites and browser extensions to use compatible AI models running locally through the browser.
2. Does WebAI send user data to cloud AI services?
For an inference task performed entirely by a local model, the content does not need to be sent to a cloud AI provider. However, websites and extensions may still transmit information for other purposes, so users should review their permissions and privacy policies.
3. Which browsers support WebAI?
The initial early preview is available in Opera One on supported 64-bit Windows PCs and Macs with Apple silicon. Opera says Opera GX and Opera Neon will follow.
4. Does WebAI require an AI subscription?
Opera says users do not need an AI subscription for the local AI capabilities described in its launch announcement. Users still need compatible hardware, and other services or extensions may have separate costs.
5. Can developers use WebAI to build AI-powered browser extensions?
Yes. The Prompt, Decision and Semantic Embeddings APIs can support extensions that generate text, classify information and identify relationships between content. Availability depends on the browser, model and permissions.
Looking Ahead
Opera’s early preview will show whether local AI can deliver a useful balance of privacy, performance and simplicity for everyday browsing. Support for additional models, more devices and wider browser adoption will be important to the standard’s long-term prospects.
For developers, the opportunity is to create AI features that work directly where users read, research and organize information, without depending on a cloud model for every operation. For users, the appeal is greater control over where AI runs and which sites can access it. Whether WebAI becomes a broadly adopted web standard will depend on its technical maturity, developer uptake and support beyond Opera.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
