Model comparison

SmolVLM2-2.2B-Instruct vs embeddinggemma-300m

These two publish different quantization sets; the table below has the exact sizes.

From the file· summed bytes, KV per layer

Side by side

SmolVLM2-2.2B-Instructembeddinggemma-300m
Parameters2.2B303M
Architecturellamagemma-embedding
Layers2424
Native context8,1922,048
Mixture of expertsnono
Quantizations published4010
Smallest quantization0.42 GiB0.26 GiB
Q4_K_M1.04 GiB
Licenceapache-2.0

KV cache by context

the term that decides long-context viability
ContextSmolVLM2-2.2B-Instructembeddinggemma-300mRatio
4,0960.04 GiB
8,1920.05 GiB
16,3840.08 GiB
32,7680.14 GiB
65,5360.27 GiB
131,0720.52 GiB