Model comparison
OpenCaption-2B-VL-SFT-v1.0 vs Unlimited-OCR
At Q4_K_M, OpenCaption-2B-VL-SFT-v1.0 is the smaller download — 1,107,408,032 bytes against 1,950,321,792.
From the file· summed bytes, KV per layer
Side by side
| OpenCaption-2B-VL-SFT-v1.0 | Unlimited-OCR | |
|---|---|---|
| Parameters | 2.1B | 3.3B |
| Architecture | qwen3vl | deepseek2-ocr |
| Layers | 28 | 12 |
| Native context | 262,144 | 32,768 |
| Mixture of experts | no | yes, 64 experts |
| Quantizations published | 50 | 21 |
| Smallest quantization | 0.48 GiB | 1.15 GiB |
| Q4_K_M | 1.03 GiB | 1.82 GiB |
| Licence | apache-2.0 | mit |
KV cache by context
the term that decides long-context viability
| Context | OpenCaption-2B-VL-SFT-v1.0 | Unlimited-OCR | Ratio |
|---|---|---|---|
| 4,096 | 0.44 GiB | — | — |
| 8,192 | 0.88 GiB | — | — |
| 16,384 | 1.75 GiB | — | — |
| 32,768 | 3.50 GiB | — | — |
| 65,536 | 7.00 GiB | — | — |
| 131,072 | 14.00 GiB | — | — |