Comment by davio

On a M1 MacBook Air with 8GB, I got this running Gemma 3n:

12.63 tok/sec • 860 tokens • 1.52s to first token

I'm amazed it works at all with such limited RAM