I bought a used m4 Mac mini 32g right before the m6 came out, so I'm trying hard to use it while holding back tears.
I was using gemma4-26b-a4b as a resident model because it was good in many ways, but I was looking for a smaller one because the memory was a little lacking. So I tested splash yesterday, but it was a bit ambiguous, so I was going to settle with gemma4-e4b and live like that, but after reading Daepung's post today, I will try splash again.
I need memory to be very small, so I'm going to test it with smaller bits. Before that, I'm going to test Bonsai2, which I didn't even look at before. I asked them to adjust it to 24g and here's the test schedule:
1. Bonsai 2 PQ2_0
2. Qwen3.6 IQ2_XXS
3. Qwen3.6 IQ3_XXS
4. Qwen3.8 IQ3_XXS
5. Final comparison with the current Gemma E4B
(Please subscribe and like. ㅎㅎ)
