Gemma 3n (E4B) extracting key details from a train ticket directly on CPU, on-device

Our latest Gemma 3n model pushes the boundaries of multimodal AI, natively understanding and processing information from images, audio, and more. See Gemma 3n in action 🧵↓

Using Gemma 3n’s advanced image understanding to parse a receipt to extract structured JSON data

Gemma 3n listening to a train announcement in German and instantly translating it into clear English text. Perfect for real-time multilingual applications, when you need them most!

Start building with Gemma 3n:

That’s a very practiced example. But looking forward to trying it out

Add a few biometrics and an olfactory array and you have a tricorder Building this shit

How to load it in the edge apk?

Uhhuh

That’s awesome

It is very good model I can run and finetune in my 16gb gpu. I tried it and it solved the problem for my specific use case. Other small models get lost.