What the fuck is Gemini Nano?

Besides Gemini Flash and Gemini Pro, there's Gemini Nano.

Get to know the small Gemini Nano model used on the web and on apps.

.

As a result of the Google IO Extended Bangkok 2025 with many sessions, Gemini Nano went to read it and made this post out of frustration.

.

First of all, Gemini Flash and Gemini Pro are large LLMs used on the server. If you want the user to download something large, then the user will wait. And there is a cost to use it.

Gemini Nano says it's on-device, it can be used in the web browser or in the Android app as it's smaller, it works even offline and privacy, because it's on our device, it's not sent anywhere.

It's Gemini 1.0 Nano right now.

Of course, there are two main groups of people who are active in implementation: web developers and android developers.

.

🔵 Web

There are 7 Built-in AI APIs.

- Translator API

Status: stable Chrome 138 Web & Extension

AI translation, which understands the context, we can press a button and translate.

- Language Detector API

Status: stable Chrome 138 Web & Extension

Before translating, you need to know what language this is, so you don't have to specify the source language.

- Summarizer API

Status: stable Chrome 138 Web & Extension

Select types, such as headlines, key points, such as redBus, press a single button, summarize the content of the page.

- Writer API

Status: Origin trial Web & Extension

Help us write in different ways, can choose tone and writing length, used to write content, review, email drafting.

- Rewrite API

Status: Origin trial Web & Extension

Write and not like. Not yet. Use this to adjust the language in email or make content addicted to SEO. It's a review of what we write with AI.

- Prompt API

Status: stable Chrome 138 and Origin trial Extension

We can type commands to do something on the extension, such as add this activity to the calendar, or write a prompt to edit content, as well as multimodal support like image and sound, which is done on the extension because Project Mariner.

- Proofreader API

Status: coming soon

It is expected that it is coming in Chrome 142. This is very helpful in the printing work of the content creator. When we type it wrong, it will tell us where we typed it wrong. How to fix it? This is hard to find, because it has to be enabled in the document.

If you want to try, you know you have to join the Early Preview Program.

.

🟢 Android.

There are 2, most of them ML Kit GenAI APIs, there are 4.

- Summarization: It's a summary of an article or a chat that can be discussed in less than three bullet lists.

- Proofreading: Please see if this one is grammar and where it's misspelled.

- Rewrite: Help us adjust the style and tone we want, with multiple output styles like Elaborate, Emojify, Shorten, Friendly, Professional and Rephrase.

- Image description: Help generate description for this photo, help the visually impaired and those who cannot use the screen at that time, such as driving

All of them are Beta, available from Android API level 26 and up. There are not many supported phones, mostly the latest ones and friends, such as Galazy S25, Pixel 9.

A rough implementation is to add the dependency of the ML Kit we use, then create the object of each API, then download the feature if available, and create a request and wait for the result.

With the Googlw AI Edge SDK, this one will be a bit advanced. Trial permissions are required on the web page.

.

It's all about this. I'll have time to hear back about gemma 3 for a second. I can see it on Android. 🤔

# googlegemini # programmer # Developer # IT should know # Includes IT matters