Model comparison

Unlimited-OCR vs Qwen2.5-VL-7B-Instruct

At Q4_K_M, Unlimited-OCR is the smaller download — 1,950,321,792 bytes against 4,683,072,032.

From the file· summed bytes, KV per layer

Side by side

Unlimited-OCRQwen2.5-VL-7B-Instruct
Parameters3.3B8.3B
Architecturedeepseek2-ocrqwen2vl
Layers1228
Native context32,768128,000
Mixture of expertsyes, 64 expertsno
Quantizations published2150
Smallest quantization1.15 GiB1.93 GiB
Q4_K_M1.82 GiB4.36 GiB
Licencemitapache-2.0

KV cache by context

the term that decides long-context viability
ContextUnlimited-OCRQwen2.5-VL-7B-InstructRatio
4,0960.22 GiB
8,1920.44 GiB
16,3840.88 GiB
32,7681.75 GiB
65,5363.50 GiB
131,0727.00 GiB