Model comparison
llava-llama-3-8b-v1_1-transformers vs gemma-4-31B-it
At Q4_K_M, llava-llama-3-8b-v1_1-transformers is the smaller download — 4,921,098,752 bytes against 18,323,733,440.
From the file· summed bytes, KV per layer
Side by side
| llava-llama-3-8b-v1_1-transformers | gemma-4-31B-it | |
|---|---|---|
| Parameters | 8.4B | 31.3B |
| Architecture | llama | gemma4 |
| Layers | — | 60 |
| Native context | 8,192 | 262,144 |
| Mixture of experts | no | no |
| Quantizations published | 15 | 55 |
| Smallest quantization | 3.41 GiB | 7.95 GiB |
| Q4_K_M | 4.58 GiB | 17.07 GiB |
| Licence | — | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | llava-llama-3-8b-v1_1-transformers | gemma-4-31B-it | Ratio |
|---|---|---|---|
| 4,096 | — | 1.80 GiB | — |
| 8,192 | — | 2.42 GiB | — |
| 16,384 | — | 3.67 GiB | — |
| 32,768 | — | 6.17 GiB | — |
| 65,536 | — | 11.17 GiB | — |
| 131,072 | — | 21.17 GiB | — |