If that is possible I would appreciate to have a good model (1) available for Android app,
That's a modern usable gemma 3 model:
(1) gemma-3n-E4B-it-Q4_K_M
The model file size can be nearly 4.23 GB --or, at least that is its size in other app which BTW can be not optimized: initially the download file size was 2.21Gb .
.
I tried other gemma 3 models and nothing of them works on my flagship smartphone device with 12Gb of RAM. Only working modern gemma 3 model was small model with 1B parameters and it's not smart at all.
The mentioned model works via other LLM runner app I use, so that's possible.