Model comparison

Qwen3.5-4B vs Unlimited-OCR

At Q4_K_M, Unlimited-OCR is the smaller download — 1,950,321,792 bytes against 2,707,513,696.

From the file· summed bytes, KV per layer

Side by side

Qwen3.5-4BUnlimited-OCR
Parameters4.7B3.3B
Architectureqwen35deepseek2-ocr
Layers3212
Native context262,14432,768
Mixture of expertsnoyes, 64 experts
Quantizations published5721
Smallest quantization1.42 GiB1.15 GiB
Q4_K_M2.52 GiB1.82 GiB
Licenceapache-2.0mit

KV cache by context

the term that decides long-context viability
ContextQwen3.5-4BUnlimited-OCRRatio
4,0960.13 GiB
8,1920.25 GiB
16,3840.50 GiB
32,7681.00 GiB
65,5362.00 GiB
131,0724.00 GiB