Model comparison

Nanonets-OCR-s vs Qwen3.5-0.8B

At Q4_K_M, Qwen3.5-0.8B is the smaller download — 527,502,816 bytes against 1,929,900,800. At long context the gap widens: Qwen3.5-0.8B's KV cache at 32K is 3.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Nanonets-OCR-sQwen3.5-0.8B
Parameters3.8B873M
Architectureqwen2vlqwen35
Layers3624
Native context128,000262,144
Mixture of expertsnono
Quantizations published5642
Smallest quantization0.74 GiB0.31 GiB
Q4_K_M1.80 GiB0.49 GiB
Licenceapache-2.0

KV cache by context

the term that decides long-context viability
ContextNanonets-OCR-sQwen3.5-0.8BRatio
4,0960.14 GiB0.05 GiB3.00×
8,1920.28 GiB0.09 GiB3.00×
16,3840.56 GiB0.19 GiB3.00×
32,7681.13 GiB0.38 GiB3.00×
65,5362.25 GiB0.75 GiB3.00×
131,0724.50 GiB1.50 GiB3.00×