Model comparison
bella-bartender-v2 vs gemma-4-E4B-it
At Q4_K_M, gemma-4-E4B-it is the smaller download — 4,977,171,584 bytes against 5,761,059,584. At long context the gap widens: gemma-4-E4B-it's KV cache at 32K is 11.8× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| bella-bartender-v2 | gemma-4-E4B-it | |
|---|---|---|
| Parameters | 9.2B | 8.0B |
| Architecture | gemma2 | gemma4 |
| Layers | 42 | 42 |
| Native context | 8,192 | 131,072 |
| Mixture of experts | no | no |
| Quantizations published | 36 | 36 |
| Smallest quantization | 2.22 GiB | 3.30 GiB |
| Q4_K_M | 5.37 GiB | 4.64 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | bella-bartender-v2 | gemma-4-E4B-it | Ratio |
|---|---|---|---|
| 4,096 | 1.31 GiB | 0.12 GiB | 10.67× |
| 8,192 | 2.05 GiB | 0.18 GiB | 11.54× |
| 16,384 | 3.36 GiB | 0.29 GiB | 11.71× |
| 32,768 | 5.99 GiB | 0.51 GiB | 11.84× |
| 65,536 | 11.24 GiB | 0.94 GiB | 11.91× |
| 131,072 | 21.74 GiB | 1.82 GiB | 11.95× |