Model comparison

TinyMistral-248M-v2-Instruct vs SmolLM2-135M-Instruct

At Q4_K_M, SmolLM2-135M-Instruct is the smaller download — 105,454,144 bytes against 155,668,768.

From the file· summed bytes, KV per layer

Side by side

TinyMistral-248M-v2-InstructSmolLM2-135M-Instruct
Parameters248M135M
Architecturellamallama
Layers1230
Native context32,7688,192
Mixture of expertsnono
Quantizations published641
Smallest quantization0.10 GiB0.08 GiB
Q4_K_M0.14 GiB0.10 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextTinyMistral-248M-v2-InstructSmolLM2-135M-InstructRatio
4,0960.09 GiB
8,1920.18 GiB
16,3840.35 GiB
32,7680.70 GiB
65,5361.41 GiB
131,0722.81 GiB