Model comparison
Qwen3.5-4B vs Nanonets-OCR-s
At Q4_K_M, Nanonets-OCR-s is the smaller download — 1,929,900,800 bytes against 2,707,513,696.
From the file· summed bytes, KV per layer
Side by side
| Qwen3.5-4B | Nanonets-OCR-s | |
|---|---|---|
| Parameters | 4.7B | 3.8B |
| Architecture | qwen35 | qwen2vl |
| Layers | 32 | 36 |
| Native context | 262,144 | 128,000 |
| Mixture of experts | no | no |
| Quantizations published | 57 | 56 |
| Smallest quantization | 1.42 GiB | 0.74 GiB |
| Q4_K_M | 2.52 GiB | 1.80 GiB |
| Licence | apache-2.0 | — |
KV cache by context
the term that decides long-context viability
| Context | Qwen3.5-4B | Nanonets-OCR-s | Ratio |
|---|---|---|---|
| 4,096 | 0.13 GiB | 0.14 GiB | 1.12× |
| 8,192 | 0.25 GiB | 0.28 GiB | 1.12× |
| 16,384 | 0.50 GiB | 0.56 GiB | 1.12× |
| 32,768 | 1.00 GiB | 1.13 GiB | 1.12× |
| 65,536 | 2.00 GiB | 2.25 GiB | 1.12× |
| 131,072 | 4.00 GiB | 4.50 GiB | 1.12× |