Model comparison
granite-4.0-h-tiny vs Qwen3-VL-8B-Instruct-abliterated-v1
At Q4_K_M, granite-4.0-h-tiny is the smaller download — 4,230,975,936 bytes against 5,027,785,792.
From the file· summed bytes, KV per layer
Side by side
| granite-4.0-h-tiny | Qwen3-VL-8B-Instruct-abliterated-v1 | |
|---|---|---|
| Parameters | 6.9B | 8.8B |
| Architecture | granitehybrid | qwen3vl |
| Layers | 40 | — |
| Native context | 131,072 | — |
| Mixture of experts | yes, 64 experts | no |
| Quantizations published | 39 | 48 |
| Smallest quantization | 1.81 GiB | 1.97 GiB |
| Q4_K_M | 3.94 GiB | 4.68 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | granite-4.0-h-tiny | Qwen3-VL-8B-Instruct-abliterated-v1 | Ratio |
|---|---|---|---|
| 4,096 | 0.03 GiB | — | — |
| 8,192 | 0.06 GiB | — | — |
| 16,384 | 0.13 GiB | — | — |
| 32,768 | 0.25 GiB | — | — |
| 65,536 | 0.50 GiB | — | — |
| 131,072 | 1.00 GiB | — | — |