Model comparison
Fallen-Gemma3-27B-v1 vs gemma-4-E4B-it
At Q4_K_M, gemma-4-E4B-it is the smaller download — 4,977,171,584 bytes against 16,546,688,704. At long context the gap widens: gemma-4-E4B-it's KV cache at 32K is 4.6× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| Fallen-Gemma3-27B-v1 | gemma-4-E4B-it | |
|---|---|---|
| Parameters | 27.4B | 8.0B |
| Architecture | gemma3 | gemma4 |
| Layers | 48 | 42 |
| Native context | — | 131,072 |
| Mixture of experts | no | no |
| Quantizations published | 25 | 36 |
| Smallest quantization | 7.86 GiB | 3.30 GiB |
| Q4_K_M | 15.41 GiB | 4.64 GiB |
| Licence | other | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Fallen-Gemma3-27B-v1 | gemma-4-E4B-it | Ratio |
|---|---|---|---|
| 4,096 | 0.67 GiB | 0.12 GiB | 5.48× |
| 8,192 | 0.91 GiB | 0.18 GiB | 5.11× |
| 16,384 | 1.38 GiB | 0.29 GiB | 4.80× |
| 32,768 | 2.31 GiB | 0.51 GiB | 4.58× |
| 65,536 | 4.19 GiB | 0.94 GiB | 4.44× |
| 131,072 | 7.94 GiB | 1.82 GiB | 4.37× |