Model comparison
gemma-4-12B-it vs llava-llama-3-8b-v1_1-transformers
At Q4_K_M, llava-llama-3-8b-v1_1-transformers is the smaller download — 4,921,098,752 bytes against 7,121,861,440.
From the file· summed bytes, KV per layer
Side by side
| gemma-4-12B-it | llava-llama-3-8b-v1_1-transformers | |
|---|---|---|
| Parameters | 12.0B | 8.4B |
| Architecture | gemma4 | llama |
| Layers | 48 | — |
| Native context | 262,144 | 8,192 |
| Mixture of experts | no | no |
| Quantizations published | 59 | 15 |
| Smallest quantization | 3.92 GiB | 3.41 GiB |
| Q4_K_M | 6.63 GiB | 4.58 GiB |
| Licence | apache-2.0 | — |
KV cache by context
the term that decides long-context viability
| Context | gemma-4-12B-it | llava-llama-3-8b-v1_1-transformers | Ratio |
|---|---|---|---|
| 4,096 | 0.72 GiB | — | — |
| 8,192 | 0.97 GiB | — | — |
| 16,384 | 1.47 GiB | — | — |
| 32,768 | 2.47 GiB | — | — |
| 65,536 | 4.47 GiB | — | — |
| 131,072 | 8.47 GiB | — | — |