Opera is introducing a new web standard called WebAI that enables users to run AI models directly in their browser without needing cloud servers. The feature supports a variety of open-weight models — including Google’s Gemma and Apple’s on-device foundation model on Macs — but Opera emphasizes the standard isn’t limited to any single model family. WebAI was built on the company’s earlier work to bring locally executed AI into its browser offerings.
The company is letting users test WebAI early via its Opera One browser. Versions of the standard for Opera GX, which is tailored for gamers and streamers, and Opera Neon, which uses AI agents to automate tasks, are expected to roll out soon. Use of the feature requires opting in; it’s not yet enabled by default.
Why WebAI matters
One key benefit of open-weight models is that they’re free to download and use, and their underlying model weights are publicly accessible. That gives developers greater flexibility to adjust behavior or performance compared to fully proprietary or locked models. Opera points out that open-weight models often have restrictions, but are still contrastable with closed-weight, fully proprietary systems.
Privacy is another major advantage. Because everything runs locally on a user’s own device, personal data doesn’t need to be sent to external servers, and there’s no need to maintain cloud AI subscriptions. Users retain full control over how the local AI model processes information.
Trade-offs and challenges
Running AI locally isn’t without costs. The computational demands fall on the user’s hardware, which can impact performance depending on device capabilities. Opera acknowledges these challenges — the performance of any AI model under WebAI will vary with the user’s machine.
Opera also points out that while local deployment improves privacy, it doesn’t erase all risks. There are still possibilities of vulnerabilities or limits in how private or secure a locally running model can be, especially if the software environment or the models themselves are compromised.
The standard supports several key functions: text generation, content classification, and analysis of relationships among pieces of information — all without needing to go over a network. Because Opera has emphasized privacy in past products, WebAI fits into its ongoing strategy of offering tools that grant users more control over their data.
This development arrives amid growing concern over how AI systems handle sensitive user data. Many generative AI tools require cloud access and continuous transmission of personal or proprietary content to remote servers. WebAI offers a model where those exchanges can happen fully on-device, which could redraw expectations around both AI privacy and performance.
What to watch: how well the performance holds up on lower-end hardware, how well Opera secures its local model ecosystem, and whether other browser makers adopt similar standards. If WebAI is successful, it could push the industry toward decentralizing AI deployment in ways that lessen dependency on large cloud providers.
