Opera launches WebAI, a proposed web standard for running AI locally in the browser

What WebAI is

Opera announced WebAI on 27 October 2026, describing it as a bigger push into local AI and a central part of its browsers. WebAI lets websites and extensions use AI that runs locally in the browser, rather than sending data to a cloud service.

WebAI is a proposed web standard, backed by companies including Google and Microsoft. Opera's implementation lets developers use a range of AI models directly in the browser, including open-weight models running on Opera's inference engine, and compatible models already on the user's device.

The APIs

WebAI introduces three APIs: Prompt, Decision and Semantic Embeddings. With them, developers can generate text, classify information and find relationships between pieces of content — building blocks for tools that do more than answer questions in a chat window.

Opera frames this as a shift from AI as a chat window to AI that works with the pages a user is on. The company says WebAI lets independent developers and small teams create their own browser capabilities instead of waiting for software companies to build them.

Availability and privacy

WebAI is opt-in: nothing is downloaded until the user turns it on in Settings, and it can be turned off at any time. The early preview is available on Windows PC (64-bit, Intel/AMD) and Macs with Apple silicon; it is not available on Intel-based Macs, Linux or Windows on Arm.

Running models locally has privacy benefits because user data does not leave the device. The trade-off is that on-device models are typically smaller and less capable than cloud models, so WebAI works best for tasks where local speed and privacy matter more than frontier-level reasoning.

What to watch

Watch whether the standard gains adoption beyond Opera and gets formalized in a standards body, since a standard that only one vendor ships is not a standard.

Watch which model providers partner with Opera and whether Apple's on-device foundation model, mentioned in coverage of the launch, becomes available through WebAI on Apple silicon Macs.

Watch how quickly competitors respond in the same tier or standard, and whether the early previews translate into real developer adoption over the next quarter.

The decisive test will be whether real developers ship real products on these foundations and whether the economics of local AI hold up against cloud inference at scale.

Why browsers are the battleground

Browsers are becoming the default runtime for AI because they are where most people and businesses already do their work. Opera's WebAI is a bet that the future of browser AI is local: models that run on the device, work with the pages the user is viewing, and do not require an API key or a cloud round-trip.

The backing from Google and Microsoft matters because those two companies control the other major browser engines. A standard they endorse has a much better chance of being adopted beyond Opera than a single-vendor proposal. For developers, that would mean writing once and running the same local AI APIs in any compliant browser.

Early-adopter feedback will matter more than benchmark claims: developers testing the previews will quickly find what the models and APIs can and cannot do, and that feedback will shape both products.

The limits

WebAI in its preview form is limited to Windows PC (Intel/AMD) and Apple silicon Macs, with no Linux or Windows on Arm support yet. The models that run locally are smaller than frontier cloud models, so capability will vary by task.

There are also open questions about how models are distributed and updated, how extensions access the APIs securely, and whether the standard will include any kind of model marketplace or trust mechanism. Opera has not yet published full details of the specification.

What the APIs do

The Prompt API generates text from a prompt, the Decision API classifies inputs into categories, and the Semantic Embeddings API turns content into vectors that can be compared for similarity and relationships. Together these cover the three most common ways applications use AI today: generation, classification and retrieval.

Opera's framing is that these are building blocks: an email client could classify incoming mail (Decision), a note app could find related notes (Embeddings), and a writing tool could draft or rewrite (Prompt) — all without any of the data leaving the device.

The bigger picture

WebAI joins a pattern of efforts to move AI computation closer to users: Liquid's LFM3 ships open-weight models sized for phones, Apple ships on-device foundation models, and Google and Microsoft are backing WebAI. The browser is the natural place for this because it is already the universal application runtime.

If the standard gets adopted, local AI in the browser could reduce dependence on cloud AI providers, lower marginal costs of AI features, and change how much data AI applications collect. Those are big shifts to watch from a standard that is currently only an early preview in Opera.