Model comparison

v6-Finch-14B-HF vs gemma-4-E4B-it

At Q4_K_M, gemma-4-E4B-it is the smaller download — 4,977,171,584 bytes against 8,767,898,176. At long context the gap widens: gemma-4-E4B-it's KV cache at 32K is 60.3× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

v6-Finch-14B-HFgemma-4-E4B-it
Parameters14.1B8.0B
Architecturerwkv6gemma4
Layers6142
Native context131,072
Mixture of expertsnono
Quantizations published2036
Smallest quantization5.10 GiB3.30 GiB
Q4_K_M8.17 GiB4.64 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
Contextv6-Finch-14B-HFgemma-4-E4B-itRatio
4,0963.81 GiB0.12 GiB30.98×
8,1927.63 GiB0.18 GiB42.90×
16,38415.25 GiB0.29 GiB53.12×
32,76830.50 GiB0.51 GiB60.29×
65,53661.00 GiB0.94 GiB64.66×
131,072122.00 GiB1.82 GiB67.09×
v6-Finch-14B-HF vs gemma-4-E4B-it — size, memory and hardware fit — ossmodeldb