Opera previews WebAI for local browser AI

Opera describes WebAI as a proposed web standard backed by companies including Google and Microsoft.[1] It lets websites and browser extensions use AI hosted through the browser, while models run locally on the user’s device rather than relying on the cloud.[1][2] Opera says its implementation is model-agnostic and built around open models.[1] At launch, it supports Google’s Gemma and, on Macs, Apple’s on-device foundation model.[1] Opera lists text generation, content classification and identifying relationships between information as supported tasks.[1]

The first WebAI capabilities are available as an early preview in Opera One; Opera says Opera GX and Opera Neon will follow.[1] The preview runs on 64-bit Windows PCs with Intel or AMD processors and on Macs with Apple silicon.[1] Opera says it is not available on Intel-based Macs, Linux or Windows on Arm.[2] Performance depends on the user’s hardware, so teams evaluating a browser-based feature should account for the devices their users have.[1]

WebAI is not enabled by default: Opera says users must turn it on in Settings before anything is downloaded, and can turn it off at any time.[2] Site-specific permissions let users decide which websites can access local AI.[1] Opera says information handled by local AI stays on the user’s device and that users do not need AI subscriptions.[1] Those controls and conditions matter for teams deciding whether a feature should rely on local processing; the preview’s availability and performance still depend on device compatibility and hardware.[1][2]

Opera says WebAI gives developers a way to add AI features without running AI infrastructure or paying for every request.[1] Its blog says developers do not need a separate AI app, local server or API key.[2] The standard introduces Prompt, Decision and Semantic Embeddings APIs.[2] Opera’s team also made six experimental extensions to illustrate possible projects, but says these are intended to spark ideas and are not Opera features.[2] Opera says it first offered built-in access to local AI models in an experimental developer build in 2024.[1]