Model comparison
PaddleOCR-VL-1.6 vs SmolLM2-135M-Instruct
At Q4_K_M, SmolLM2-135M-Instruct is the smaller download — 105,454,144 bytes against 300,219,456. At long context the gap widens: PaddleOCR-VL-1.6's KV cache at 32K is 1.3× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| PaddleOCR-VL-1.6 | SmolLM2-135M-Instruct | |
|---|---|---|
| Parameters | 959M | 135M |
| Architecture | paddleocr | llama |
| Layers | 18 | 30 |
| Native context | 131,072 | 8,192 |
| Mixture of experts | no | no |
| Quantizations published | 36 | 41 |
| Smallest quantization | 0.15 GiB | 0.08 GiB |
| Q4_K_M | 0.28 GiB | 0.10 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | PaddleOCR-VL-1.6 | SmolLM2-135M-Instruct | Ratio |
|---|---|---|---|
| 4,096 | 0.07 GiB | 0.09 GiB | 1.25× |
| 8,192 | 0.14 GiB | 0.18 GiB | 1.25× |
| 16,384 | 0.28 GiB | 0.35 GiB | 1.25× |
| 32,768 | 0.56 GiB | 0.70 GiB | 1.25× |
| 65,536 | 1.13 GiB | 1.41 GiB | 1.25× |
| 131,072 | 2.25 GiB | 2.81 GiB | 1.25× |