Model comparison

HunyuanOCR vs embeddinggemma-300m

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: embeddinggemma-300m's KV cache at 32K is 5.8× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

HunyuanOCRembeddinggemma-300m
Parameters1.1B303M
Architecturehunyuan_vlgemma-embedding
Layers2424
Native context131,0722,048
Mixture of expertsyes, 1 expertsno
Quantizations published2510
Smallest quantization0.24 GiB0.26 GiB
Q4_K_M0.33 GiB
Licence

KV cache by context

the term that decides long-context viability
ContextHunyuanOCRembeddinggemma-300mRatio
4,0960.11 GiB0.04 GiB3.00×
8,1920.21 GiB0.05 GiB4.15×
16,3840.42 GiB0.08 GiB5.14×
32,7680.84 GiB0.14 GiB5.84×
65,5361.69 GiB0.27 GiB6.26×
131,0723.38 GiB0.52 GiB6.50×