Opera has introduced WebAI, an interface that lets websites call AI models running locally inside its browser. The early preview is available as an opt-in feature in Opera One, with support planned for Opera GX and the agent-focused Opera Neon.
WebAI can generate text, classify content and represent relationships between pieces of information without sending those inputs to a remote model. Opera says the interface is not tied to one provider. It can use open-weight systems such as Google’s Gemma and, on Macs, Apple’s on-device foundation model.
Local execution changes the economics for developers because a site can add AI functions without operating inference servers or paying for every model request. It can also keep sensitive input on the user’s computer, reducing exposure to a cloud provider. That does not remove every privacy risk: websites still need clear permissions, and locally generated output can still be stored or transmitted by application code.
The trade-off is hardware. Model speed, memory use and capability depend on the person’s computer, and sustained inference can consume battery and other resources. WebAI is currently an Opera proposal and early preview, not a broadly adopted web standard. Developers considering it will need fallbacks for unsupported browsers and lower-powered devices while the interface and model availability mature.