Google introduced a number of thrilling Gemini AI improvements at I/O 2025, considerably enhancing Gemini’s talents throughout the board and bringing the AI to extra of its apps. However Gemini is a proprietary AI mannequin that you should use solely with a Google account. The AI Mode in Google Search could be an exception, but it surely doesn’t provide full entry to what Gemini can do.
Additionally, Gemini requires an web connection more often than not. The AI handles information processing within the cloud for many duties, particularly complicated ones. Gemini Nano, nevertheless, is a mannequin designed to carry out on-device duties on smartphones.
If you happen to’re searching for Gemini tech in an open-source format or desire on-device AI, try Gemma 3n. This Google AI mannequin was introduced at I/O 2025. It’s constructed on the identical tech as Gemini Nano and is designed particularly to run on-device.
Google stated in a weblog put up that it engineered a brand new “cutting-edge structure” for the subsequent era of on-device AI. The structure was optimized for “lightning-fast, multimodal AI, enabling really private and personal experiences straight in your system,” in partnership with Qualcomm, MediaTek, and Samsung.
Gemma 3n is the primary open mannequin constructed on this new structure, and builders can begin utilizing it straight away. That’s the important thing element right here. It is advisable to know what you’re doing to make use of Gemma 3n. It’s not an app you run on an iPhone or Android like Gemini. You’ll have to put in it in your laptop or use it to develop your individual apps that ship on-device AI.
The partnerships talked about above counsel that loads of third events will likely be inquisitive about utilizing Gemini-like AI tech through Gemma 3n to supply on-device AI experiences on telephones, tablets, and laptops.
Google additionally says Gemma 3n could be very environment friendly, permitting the AI to function with simply 2GB or 3GB of RAM. It’s additionally very quick in comparison with different proprietary and open fashions. The Chatbot Enviornment Elo scores place it alongside Anthropic’s Claude 3.7 Sonnet.
So, what can Gemma 3n do? As you’ll see within the demo beneath, it’s a multimodal mannequin like Gemini. Which means it may possibly deal with textual content and voice inputs and perceive photos, whether or not on the display screen or by means of dwell video from the digital camera.
Gemma 3n can acknowledge written textual content, carry out translations, and reply complicated questions, together with fixing math issues in real-time.
Google’s Gemma 3n expertise resembles what Gemini and Gemini Dwell provide on cell gadgets. The distinction is that Gemma 3n is optimized to run on-device and is accessible for integration into apps or working methods, quite than being a standalone app from Google.
For instance, some Android distributors may use Gemma 3n to supply customized on-device options that don’t rely on Google. Builders may need to add superior AI assist to their apps, and open fashions like Gemma 3n may very well be a fantastic match.
Then there’s the wave of AI good glasses anticipated to hit shops within the coming years. Some will come from Google and its companions and run Gemini. OpenAI can be working by itself ChatGPT {hardware}, and it’s attainable they’re growing glasses too. Different distributors may desire an open-source mannequin to ship high-end AI options, and Gemma 3n may very well be their alternative.
Put merely, fashions like Gemma 3n may very well be key to the way forward for private, on-device AI experiences, even when they don’t get as a lot consideration because the proprietary AIs they’re constructed on.
