Opera WebAI: Run Local AI in Your Browser Without an API Key
Opera WebAI lets websites and extensions use local AI models without cloud services, API keys or subscriptions. On Macs, Opera One uses Apple's model.
👁 113 reads
Opera has added WebAI to its browsers, giving websites and extensions access to AI models that run on your computer. They can work without the cloud, an API key or a subscription. On Tuesday, October 27, 9to5Mac reported a notable detail: on Macs, Opera One uses Apple’s model, which is already built into the system. Here’s how WebAI works, what it runs on and how you can enable it.
What is Opera WebAI?
WebAI is a proposed web standard. Opera says it is supported by companies including Google and Microsoft. It lets a website or extension ask the browser to use a local AI model instead of sending text to an OpenAI or Google server. Opera initially offers three interfaces:
- Prompt API - text generation,
- Decision API - content classification,
- Semantic Embeddings API - finding similarities between passages of text.
The implementation is not tied to a single model. It initially runs Google’s Gemma, and on Macs it can also use Apple’s local foundation model, the same one that powers Apple Intelligence. Read our guide to Google’s open Gemma models.
What devices support WebAI?
WebAI is in an early testing phase. It works on 64-bit Windows PCs with Intel or AMD processors and on Macs with Apple processors. It does not work on Intel-based Macs, Linux or Windows on ARM. Opera says performance depends on your hardware because the model runs on your processor and graphics card.
How to enable WebAI in Opera
- Install or update Opera One.
- Go to Settings, then Opera AI - or enter
opera://settings/ai- and enable the Web AI API option. The browser will then download a local model, which may take a few minutes. - Install one of Opera’s example extensions.
The feature is disabled by default. Nothing downloads until you enable it. You can also choose which websites have access to the local model.
Six experimental Opera extensions
Opera has built six experimental extensions to demonstrate what developers can make. Project Lens colours article paragraphs according to how closely they match your project. Got It? creates a quick quiz based on what you’ve read. Semantic Art turns an article into a moving map of sentences, placing similar sentences next to one another. Other examples are made for fun, such as two ASCII cats improvising a scene on a given topic. Opera stresses that these are examples, not built-in browser features.
Opera also showed HTML-in-Canvas: real, clickable websites embedded in 2D and 3D scenes. Buttons and forms still work even when a page moves around inside a three-dimensional room.
What WebAI could mean for developers and privacy
The key question is who pays for the computing. Today, every AI feature on a website means an API bill for its creator or a subscription for its users. WebAI shifts the cost to the user’s computer. That’s a major difference for small extension developers: they can add AI without running a server or paying for every request. It could also be better for privacy because text from a webpage stays on your computer.
There is one catch: WebAI is still a proposed standard. Until Chrome and Edge implement the same interface, an extension built for WebAI will work only in Opera. Small local models are also much less capable than cloud models. They may be enough for summarising or classification, but not for serious writing. If you want full control over your model, you can use Ollama or explore the other tools in our guide to local AI.
Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →
