The browser vendor Opera has introduced a new web standard that allows users to run AI models directly in their browser without relying on the cloud.
The new standard, called WebAI, is designed to be compatible with multiple AI models. Opera mentioned Google’s Gemma as an example—a publicly accessible model first introduced by Google two years ago. The company stated that on Mac computers, WebAI also supports Apple’s on-device foundational models.
However, Opera stated that WebAI is not tied to any specific model but is designed for a broad range of open-weight AI systems. WebAI stems from Opera’s earlier work on bringing locally running models into the browser.
Opera is offering these new AI features as an early preview in its flagship browser, Opera One. The company states that the same features will soon be available in Opera GX, the browser tailored for gamers and streamers, as well as Opera Neon, the browser that uses AI agents to perform tasks. According to the company, users must actively opt in to use this feature.
Open-weight models offer several benefits to users. First, these models are often available for free download and use. The underlying model weights are also made public, giving developers greater freedom to customize how the model operates. Although distinct from fully open-source software, open-weight models still come with certain restrictions.
The company states that privacy is another advantage: users can utilize AI features without sending personal information to the company’s servers or paying for cloud-based AI subscriptions.
Opera stated in its announcement: "Developers can add AI features without running AI infrastructure or paying for each request. Users do not need an AI subscription, and locally processed AI data remains on their own devices, keeping them in control of AI on their computers."
The company says that WebAI also allows anyone to generate text, classify content, and understand relationships between pieces of information directly within their browser.
As AI assistants become more powerful and gain access to personal information, privacy concerns are growing. Users often share large amounts of personal data when using AI chatbots, sometimes without understanding how this data is stored or used.
Running models locally provides an alternative, as processing occurs on the user’s own device rather than the company’s cloud servers. While this doesn’t eliminate all privacy risks, it does give users greater control. This explains why Opera has long positioned digital privacy as one of its key selling points.
However, this approach involves trade-offs. Running AI models locally places higher demands on the user’s computer and may impact performance. Opera notes that model performance will depend on the device’s hardware.