Opera WebAI Brings Free Local AI to Your Browser — No Subscription, No Cloud

On October 27, 2026, Opera gave its two-year local-AI project a name and a public preview: WebAI, now rolling out as an opt-in early preview in Opera One, with Opera GX and Opera Neon to follow. The pitch is simple — local AI models, running in the browser, with no subscription, no account and no data sent to a cloud AI provider. The reality has a few asterisks, worth reading before you turn it on.

WebAI is not another chatbot sidebar. It is a proposed web standard, backed by companies including Google and Microsoft according to Opera, that lets websites and browser extensions call AI models hosted through the browser itself. The model runs on your computer, so the text it processes does not need to travel to a company's servers. Opera calls its implementation model-agnostic and built around open models: at launch, Google's open-weight Gemma models, and on Macs, Apple's on-device foundation model.

What WebAI actually does

Strip away the standards language and WebAI gives a web page three abilities, all running locally: generate text, classify content, and work out how pieces of information relate to each other. Opera's blog names three developer interfaces — a Prompt API, a Decision API and a Semantic Embeddings API — while Opera's developer playground lists eight, adding Summarizer, Writer, Rewriter, Translator and Proofreader.

The use cases Opera keeps returning to are deliberately mundane: an extension that sorts a messy pile of tabs, flags clickbait, or spots relevant job posts. None of this needs a data centre. All of it needs a computer with enough spare power to run a small model — the first real catch.

How to turn it on

WebAI is off by default. Nothing is downloaded until you switch it on, and you can switch it off again at any time. The steps, straight from Opera's own instructions:

1. Download or update Opera One.
2. Go to Settings, then Opera AI — or open opera://settings/ai — and turn on "Enable Web AI API." The browser then downloads a compatible local model, which Opera says can take a few minutes.
3. Install one of the example extensions and open it from your extensions menu.

Two permission layers sit on top of that switch: WebAI is opt-in at the browser level, and each website still has to ask for your permission before it can use the local model.

Will it run on your machine?

The preview supports 64-bit Windows PCs with Intel or AMD processors and Macs with Apple silicon. It does not run on Intel-based Macs, Linux, or Windows on Arm. Opera has not published a minimum memory figure or the size of the model download — only that performance depends on your hardware and that local AI performance is, in its own words, "still improving." On a thin ultrabook, the free local model may feel sluggish next to a cloud model running on a server farm, and there is no published benchmark to compare the two.

What it costs — the fine print

The price is the easiest part of this story to verify: there is not one. Opera One is free to download, and WebAI needs no AI subscription, no account, no API key and no per-request payment — the model runs on hardware you already own, so there is nobody to bill you. For developers, adding an AI feature no longer means running AI infrastructure or paying a provider for every request: a solo developer can ship an AI feature whose marginal cost is zero, because the user's laptop pays the compute bill.

The fine print: "free" means free of money, not free of cost. The compute lands on your device — battery, fan noise, and storage for a download whose size Opera has not published. And the models are capable small open-weight models, not frontier models: on the hardest reasoning, the cloud still wins.

The privacy story — and its limits

Privacy is the headline reason Opera gives for doing this at all. When the model runs on your laptop, the text you feed it does not go to a cloud AI provider — a real difference from a cloud chatbot, where every prompt is a round trip to someone else's server. TechCrunch's coverage makes the same point: local processing gives users more control over personal information AI assistants increasingly touch.

But local inference is not the same as total privacy, and Opera does not claim otherwise. WebAI controls what the model does with your text; it does not control what a website does afterwards. A site could still send the model's output — or other page data — to its own servers, which is why the per-site permission layer exists. WebAI removes the AI provider from the data's journey. It does not remove the website.

WebAI vs Chrome's built-in AI vs cloud AI

Opera is not the first browser to try this. Chrome has its own built-in AI, which runs Google's Gemini Nano on-device, and Opera's implementation appears to share the same developer interface — Opera's Prompt playground links to Google's Prompt API documentation. The models differ, so results for the same prompt can differ. Independent coverage of Chrome's built-in AI notes it asks for 22 GB of free storage plus a capable GPU or 16 GB of RAM; Opera has published no equivalent figure for WebAI. The comparison, as of launch day:

The trade-off is plain: WebAI and Chrome's built-in AI win on cost and on keeping data local; cloud AI wins on raw capability. Opera's differentiator is not the model — Gemma is Google's — it is the standard. WebAI is proposed as a web standard backed by Google and Microsoft, which, if it sticks, means a site could one day call local AI the same way in any browser that supports it. That is a bigger ambition than a feature, and the part most likely to move slowly.

What developers can build with it today

To show what is possible, Opera's team built six experimental extensions — examples to spark ideas, Opera stresses, not Opera features:

The six: Project Lens, which colours any article by how much each paragraph matters to your project; Got It?, which writes a quiz about the article you just read; Questionable Intelligence, which asks playful questions and draws the verdict as a chart; Semantic Art, which turns an article into moving art; RecordIn Alchemy, which turns any article into an insufferably inspirational social media post; and C·I·A·O (Cat Intelligence at Opera), in which two ASCII cats improvise a scene on any topic you give them.

Behind the jokes, the serious point is aimed at independent developers and small teams: page data can stay on the user's device, there are no per-request cloud costs and no AI infrastructure to run. That is the pitch that could matter most: not a new AI feature, but a new zero-cost way to ship AI features.

Also launching: HTML-in-Canvas

The same update ships a second feature: HTML-in-Canvas, which puts real, interactive web pages inside 2D and 3D scenes — menus, buttons and forms move with the scene and still work. Opera's demos include Webpage Bulldozer and Gamer Room. A natural fit for games and Opera GX, but a sideshow next to WebAI.

Pros and cons

Pros: genuinely free — no subscription, no account, no API key, no per-request cost; data processed by the AI stays on your device; opt-in with per-site permissions, and nothing downloads until you enable it; model-agnostic and built on open models; developers can ship AI features with zero cloud infrastructure.

Cons: small models only — Gemma and Apple's on-device model cannot match frontier cloud models on hard tasks; performance depends entirely on your hardware, with no published minimum specs or benchmarks; no Linux, no Windows on Arm, no Intel Macs; a website can still send the AI's output onward, so local inference is not total privacy; early preview means rough edges, and GX and Neon are still waiting.

FAQ

Is Opera WebAI really free?
Yes. Opera One is free to download, and WebAI needs no AI subscription, no account and no API key. The cost is the compute on your own device — battery, processing power and the storage for the downloaded model, whose size Opera has not published.

Which models does WebAI use?
At launch, Google's open-weight Gemma models on supported Windows PCs and Macs, and Apple's on-device foundation model on compatible Macs. Opera calls the implementation model-agnostic, so the list can grow as the preview develops.

How do I turn on WebAI in Opera One?
Update Opera One, go to Settings, then Opera AI (or open opera://settings/ai), and switch on "Enable Web AI API." The browser downloads a compatible local model, which Opera says can take a few minutes. Each website then asks for your permission before it can use the model.

Does WebAI keep my data private?
It keeps the text you process away from cloud AI providers, because the model runs on your machine. It does not control what a website does with the model's output afterwards — a site could still send results to its own servers. Per-site permissions let you decide which sites may use the local model at all.

Is WebAI the same as Chrome's built-in AI?
They appear to share the same developer interface, and both run AI on your device for free. The models differ: Chrome uses Gemini Nano, while Opera uses Gemma or Apple's model. Results for the same prompt can therefore differ between the two browsers.

Key takeaways

• Opera launched WebAI on October 27, 2026, as an opt-in early preview in Opera One, with Opera GX and Opera Neon to follow.
• Websites and extensions can use local AI — Google's Gemma, or Apple's on-device model on Macs — with no subscription, account or per-request fee.
• Turning it on takes one switch in Settings, then a model download that Opera says takes a few minutes; each site still needs your permission.
• The preview runs only on 64-bit Windows (Intel/AMD) and Apple-silicon Macs, and Opera has published no minimum specs or benchmarks.
• Local inference keeps your text off cloud AI servers, but a website can still send the AI's output onward — the per-site permission is the real privacy control.

Opera's bet is that a lot of what people want from AI does not need a data centre — just a laptop, a small open model, and a browser willing to get out of the way. On October 27, that bet became something you can actually download.

Sources: Opera Newsroom — Opera adds local AI to its browsers (official, October 27, 2026); Opera Blog — WebAI: Local AI for the web, now in Opera (official); TechCrunch — Opera introduces a new web standard to let users deploy AI locally in their browser; Unite.AI — Opera Launches WebAI, Bringing Local AI Models to Its Browsers; SaveDelete — Opera WebAI: How Local AI in the Browser Works; Digital Trends — Opera wants websites to run AI on your laptop instead of sending your data to the cloud

Related reading on AI Core Updates: Perplexity Computer Goes Fully Local on Mac: Setup, the 36GB Hardware Bar, and the Credit Catch; Recursive Flint: Free 40B Laptop Coding Model Built Largely by AI