Gemma 3 Surprised Me

I was surprised by Gemma 3’s capabilities. I tried its 4-bit quantized 12B model through Ollama.

I’ve only talked with it a little so far, but it felt similar to GPT-4o. I wouldn’t have believed this kind of progress was possible two years ago.

Now, I’m starting to want an M4 for AI tasks!

This text was also edited by Gemma.