If you own a Pixel 11, Gemini Nano 4 is now something you can try yourself instead of just read about. A hidden switch in Developer options unlocks Google’s newest on-device AI models, and you don’t need a beta-tester sign-up anymore. The catch: it’s a preview, and it behaves like one.
Here’s the short version before the details.
- Gemini Nano 4 is Google’s next on-device model, built on Gemma 4.
- Seven phones are listed: four Pixel 11 models and three Galaxy Z foldables.
- Pixel 11 owners can turn on the preview under Settings > System > Developer options > AICore Settings.
- Google hasn’t given a public release date for the final version.
That list is enough to decide whether this post applies to you. The rest covers what the model is, how to switch it on, and whether you should.
What Gemini Nano 4 actually is
Gemini Nano is the small model that runs directly on your phone instead of in Google’s cloud. That matters for speed, privacy, and working without a signal. Gemini Nano 4 is the next generation, and Jetstream reports it’s built on the multimodal Gemma 4 family that Google announced in April.
It comes in two flavors. E4B is the fuller model for harder reasoning, and E2B is tuned for speed with lower latency. Jetstream also says Google claims faster inference and lower battery use than the current model, plus support for more than 140 languages. Treat those numbers as claims until independent testing backs them up.
Why should you care? On-device models answer without sending your text or voice to a server. They also keep working on a plane or in a basement. The trade-off is size: a phone can’t match a data-center model, so Nano models handle quick, focused jobs like summaries, rewrites, and voice cleanup. Heavier questions still go to the cloud.
If you’ve followed Google’s offline push, this is the engine behind it. We covered the user-facing side in our Gemini Live offline mode guide.
Which phones support Gemini Nano 4
Google quietly added the supported devices to its developer documentation, according to 9to5Google. Only seven phones are on the list so far.
| Brand | Phones listed | Preview switch confirmed? |
|---|---|---|
| Pixel 11, 11 Pro, 11 Pro XL, 11 Pro Fold | Yes, per Jetstream (Pixel 11) | |
| Samsung | Galaxy Z Flip 8, Z Fold 8, Z Fold 8 Ultra | Not reported |
Older phones like the Pixel 10 and Galaxy S26 aren’t listed. It’s also unclear whether they’ll get Gemini Intelligence features later, so don’t count on it.
How to turn on the Gemini Nano 4 preview
Until recently, you had to join a Google Group and register as an AICore beta tester. Jetstream says that’s no longer needed on the Pixel 11, because the toggle showed up around November 6.
Here’s the path it describes. You’ll need Developer options unlocked first, which means tapping the build number in About phone seven times.
- Open Settings, then System, then Developer options.
- Tap AICore Settings.
- Switch on “Enable AICore Developer Preview.”
- Look for a new AICore app in your app drawer.
Inside that app you’ll see four model options, including Full and Fast variants. Jetstream notes a built-in test screen lets you set system prompts, temperature, and thinking mode, and shows time to first token. It’s a developer tool, so expect a plain interface.
The Gemini Intelligence confusion
Gemini Nano 4 is not the same thing as Gemini Intelligence, and mixing them up is easy. Gemini Intelligence is the consumer feature set, and 9to5Google says Google’s page asks for “Nano v3 or greater,” not Nano 4 specifically.
The listed hardware bar is steep: 12GB or more of RAM, a qualified flagship chip, five OS upgrades, and six years of security updates. Features like Rambler voice typing are Pixel 11 only for now, as we explained in our Gboard Rambler post.
So flipping the developer switch won’t suddenly give you every new AI feature. It gives developers a way to test models. Consumer features will arrive in waves, and Google hasn’t published a timetable.
Should you enable it?
Google’s own warning, as relayed by Jetstream, is that preview models may be slow, inaccurate, and can affect stability, storage, and performance. That’s a fair description of any early AI build.
If you build Android apps, turn it on and test. If you just want better AI on your daily phone, wait for the stable release and use your phone normally in the meantime. Your battery and storage will thank you.
Frequently Asked Questions
Is Gemini Nano 4 available on my Pixel 10?
Not according to the current device list, which names only Pixel 11 and select Galaxy foldables. Google hasn’t said whether older phones will be added.
Does Gemini Nano 4 need an internet connection?
No, the point of an on-device model is that it runs locally. You may still need to download model files the first time.
Is the AICore developer preview safe to use?
It’s off by default and reversible, but Google warns preview models can be slow and affect device stability. Avoid it on a phone you can’t afford to have misbehave.
When will Gemini Nano 4 launch officially?
There’s no public release date yet. The models are already on the supported phones in preview form.
My take: this is a developer story that happens to be visible to everyone. Unless you write Android apps, leave the preview switch alone and wait for the features Google finishes first.
Leave a Reply