All Android app devs can now experimentally access Gemini Nano on Pixel 9

Since launch, access to Gemini Nano has been limited to Google and OEM apps, but broader experimental availability for Android app developers is now here.

Besides first-party apps, Google initially gave a “limited set of partners” access to the on-device model. Those partner applications are not yet here.

Google is now “opening up access to experiment with Gemini Nano to all Android developers with the AI Edge SDK via AICore” on the Pixel 9 series. 

Notably, this “experimental access is for development purposes” and not stable “production usage.” App developers will specifically be able to “experiment with text-to-text prompts,” while “support for more devices and modalities” is coming in the future.

The process involves joining the aicore-experimental Google group to gain access to “Android AICore (Beta)” via the Play Store.

• “Check out our documentation to get started.”
• “If you want to test your own use case, you can download our sample app for an easy way to start experimenting with Gemini Nano.”

Developers are getting access to a Gemini 1.0 Nano model with 3.25B parameters, which is referred to as “Nano 2” in the academic paper and chart below (versus 1.8B “Nano 1”):

Compared to its predecessor, the model being made available to developers today (referred to in the academic paper as “Nano 2”) delivers a substantial improvement in quality. At nearly twice the size of the predecessor (“Nano 1”), it excels in both academic benchmarks and real-world applications, offering capabilities that rival much larger models.

Specific functionality available includes:

• Rephrasing: Rephrasing and rewriting text to change the tone to be more casual or formal.
• Smart reply: Given several chat messages in a thread, suggest the next likely response.
• Proofreading: Removing spelling or grammatical errors from text.
• Summarization: Generating a summary of a long document, either as a paragraph or as bullet points.