Model comparison

granite-vision-3.2-2b vs Qwen3-VL-2B-Instruct

At Q4_K_M, Qwen3-VL-2B-Instruct is the smaller download — 1,107,409,952 bytes against 1,545,294,400. At long context the gap widens: granite-vision-3.2-2b's KV cache at 32K is 1.4× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

granite-vision-3.2-2bQwen3-VL-2B-Instruct
Parameters3.0B2.1B
Architecturegraniteqwen3vl
Layers4028
Native context131,072262,144
Mixture of expertsnono
Quantizations published3046
Smallest quantization0.83 GiB0.50 GiB
Q4_K_M1.44 GiB1.03 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
Contextgranite-vision-3.2-2bQwen3-VL-2B-InstructRatio
4,0960.31 GiB0.44 GiB1.40×
8,1920.63 GiB0.88 GiB1.40×
16,3841.25 GiB1.75 GiB1.40×
32,7682.50 GiB3.50 GiB1.40×
65,5365.00 GiB7.00 GiB1.40×
131,07210.00 GiB14.00 GiB1.40×
granite-vision-3.2-2b vs Qwen3-VL-2B-Instruct — size, memory and hardware fit — ossmodeldb