Model comparison

OvisOCR2 vs MiniCPM-V-4.6

At Q4_K_M, MiniCPM-V-4.6 is the smaller download — 529,101,504 bytes against 529,297,280.

From the file· summed bytes, KV per layer

Side by side

OvisOCR2MiniCPM-V-4.6
Parameters853M1.3B
Architectureqwen35qwen35
Layers2424
Native context262,144262,144
Mixture of expertsnono
Quantizations published4315
Smallest quantization0.35 GiB0.47 GiB
Q4_K_M0.49 GiB0.49 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextOvisOCR2MiniCPM-V-4.6Ratio
4,0960.05 GiB0.05 GiB1.00×
8,1920.09 GiB0.09 GiB1.00×
16,3840.19 GiB0.19 GiB1.00×
32,7680.38 GiB0.38 GiB1.00×
65,5360.75 GiB0.75 GiB1.00×
131,0721.50 GiB1.50 GiB1.00×