Gemma quant comparison on M5 Max MacBook Pro 128GB (subjective of course, but on variety of categories): gemma 4 leaderboard the surprising bit: Gemma 4 31B 4bit scored higher than 8bit. 91.3% vs 88.4%. not sure why: could be the template, could …
A year ago I would just read about 397B league of models. Today I can run it on my laptop. The combination of importance matrix (imatrix) with Unsloth's per-model adaptive layer quantization is what makes it all possible. But I didn't sta…
The last Llama (Scout/Maverick) was released a year ago. Since then US based releases have been super rare: Granite 3.3, GPT-OSS 20B & 120B, Nemotron 3 Nano / Super and now Gemma 4. Can't even compare to the solid Chinese open model outpu…