Model comparison

DeepSeek-OCR-2 vs Qwen3-VL-8B-Instruct-abliterated-v1

At Q4_K_M, DeepSeek-OCR-2 is the smaller download — 1,950,326,688 bytes against 5,027,785,792.

From the file· summed bytes, KV per layer

Side by side

DeepSeek-OCR-2Qwen3-VL-8B-Instruct-abliterated-v1
Parameters3.4B8.8B
Architecturedeepseek2-ocrqwen3vl
Layers12
Native context8,192
Mixture of expertsyes, 64 expertsno
Quantizations published848
Smallest quantization1.35 GiB1.97 GiB
Q4_K_M1.82 GiB4.68 GiB
Licenceapache-2.0

KV cache by context

the term that decides long-context viability
ContextDeepSeek-OCR-2Qwen3-VL-8B-Instruct-abliterated-v1Ratio
4,0960.12 GiB
8,1920.23 GiB
16,3840.47 GiB
32,7680.94 GiB
65,5361.88 GiB
131,0723.75 GiB