Any tab below showing "Model download required" can pull a model anywhere from a couple hundred MB to a few GB the first time you click "Download model" - that's a real download to your device's disk, not just a check. It only ever starts when you explicitly click that button; opening or switching tabs here never starts one on its own.
Chrome ships a set of experimental on-device AI APIs - no server involved, no text ever leaves your machine. Most of these currently require an origin trial or an enabled flag, so "not supported" below is the expected result in most browsers. This tab only checks availability - opening a tab for a "model download required" API is a separate step and won't start anything from here.
General-purpose text prompting against Chrome's on-device model (Gemini Nano). Chrome docs
Condenses long text into key points, a TL;DR, a teaser, or a headline. Chrome docs
Translates text between a language pair, entirely on-device. Shown here for English -> Spanish as a sample pair - open the Translator tab to check others. Chrome docs
Identifies the language(s) a piece of text is written in. Chrome docs
Drafts new text for a task you describe, in a chosen tone and length. Chrome docs
Revises existing text - tone, length, or format - without changing its meaning. Chrome docs
The GPU access API these models use for fast on-device inference where available. Chrome docs
A lower-level neural-network execution API some on-device models can use instead of WebGPU. Chrome docs
This page tests Chrome's experimental Built-in AI APIs - on-device text models (Prompt, Summarizer, Translator, Language Detector, Writer, Rewriter) that Chrome can run locally instead of calling a server. Nothing you type here is sent anywhere; every check and every run happens in your browser.
Most of these APIs currently require an origin trial or a Chrome flag, so seeing "Not supported in this browser" on every tab is the expected result for most visitors - the Overview tab explains what each one does regardless of whether it's available to you.
What is Chrome's "Built-in AI"?
A set of experimental browser APIs (Prompt, Summarizer, Translator, Language Detector, Writer, Rewriter) that run a text model directly on your device — Gemini Nano — instead of calling a server. This tool checks which of them your browser exposes and lets you try each one.
Does any text I type here get sent to a server?
No — every check and every prompt runs entirely in your browser, using the browser's own on-device model. This site has no involvement in and no visibility into any of it.
Why does every check say "Not supported in this browser"?
These APIs currently ship behind an origin trial or a Chrome flag (chrome://flags), so most visitors — including most Chrome users — won't have them enabled. That's expected, not a bug in this tool.
What does "Model download required" mean?
The API exists in your browser but its model hasn't been downloaded to your device yet. Downloading is a separate, explicit step in this tool — nothing downloads until you click "Download model" — and can range from a couple hundred MB to a few GB.