Model comparison

gemma-3-12b-it-qat-q4_0-unquantized vs gemma-4-E2B-it

At Q4_K_M, gemma-4-E2B-it is the smaller download — 3,106,738,272 bytes against 7,300,778,368.

From the file· summed bytes, KV per layer

Side by side

gemma-3-12b-it-qat-q4_0-unquantizedgemma-4-E2B-it
Parameters12.2B5.1B
Architecturegemma3gemma4
Layers35
Native context131,072
Mixture of expertsnono
Quantizations published4354
Smallest quantization2.85 GiB2.13 GiB
Q4_K_M6.80 GiB2.89 GiB
Licencegemmaapache-2.0

KV cache by context

the term that decides long-context viability
Contextgemma-3-12b-it-qat-q4_0-unquantizedgemma-4-E2B-itRatio
4,0960.05 GiB
8,1920.08 GiB
16,3840.14 GiB
32,7680.25 GiB
65,5360.46 GiB
131,0720.90 GiB