If you are interested in small language model to fine tune, gemma3:270m is quite interesting for its size