Model comparison

gemma-4-E4B-it-assistant vs TinyMistral-248M-v2-Instruct

At Q4_K_M, gemma-4-E4B-it-assistant is the smaller download — 77,231,200 bytes against 155,668,768.

From the file· summed bytes, KV per layer

Side by side

gemma-4-E4B-it-assistantTinyMistral-248M-v2-Instruct
Parameters79M248M
Architecturegemma4-assistantllama
Layers412
Native context131,07232,768
Mixture of expertsnono
Quantizations published96
Smallest quantization0.07 GiB0.10 GiB
Q4_K_M0.07 GiB0.14 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
Contextgemma-4-E4B-it-assistantTinyMistral-248M-v2-InstructRatio
4,0960.01 GiB
8,1920.02 GiB
16,3840.04 GiB
32,7680.07 GiB
65,5360.13 GiB
131,0720.26 GiB