Model comparison
Meta-Llama-3.2-3B vs Qwen3-VL-8B-Instruct-abliterated-v1
At Q4_K_M, Meta-Llama-3.2-3B is the smaller download — 2,019,373,888 bytes against 5,027,785,792.
From the file· summed bytes, KV per layer
Side by side
| Meta-Llama-3.2-3B | Qwen3-VL-8B-Instruct-abliterated-v1 | |
|---|---|---|
| Parameters | 3.2B | 8.8B |
| Architecture | llama | qwen3vl |
| Layers | — | — |
| Native context | — | — |
| Mixture of experts | no | no |
| Quantizations published | 5 | 48 |
| Smallest quantization | 1.88 GiB | 1.97 GiB |
| Q4_K_M | 1.88 GiB | 4.68 GiB |
| Licence | llama3 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Meta-Llama-3.2-3B | Qwen3-VL-8B-Instruct-abliterated-v1 | Ratio |
|---|---|---|---|
| 4,096 | — | — | — |
| 8,192 | — | — | — |
| 16,384 | — | — | — |
| 32,768 | — | — | — |
| 65,536 | — | — | — |
| 131,072 | — | — | — |