Model comparison
gemma-4-12B-coder-fable5-composer2.5-v1-bf16 vs gemma-4-E4B-it
These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: gemma-4-E4B-it's KV cache at 32K is 4.9× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| gemma-4-12B-coder-fable5-composer2.5-v1-bf16 | gemma-4-E4B-it | |
|---|---|---|
| Parameters | 12.0B | 8.0B |
| Architecture | gemma4 | gemma4 |
| Layers | 48 | 42 |
| Native context | 262,144 | 131,072 |
| Mixture of experts | no | no |
| Quantizations published | 24 | 36 |
| Smallest quantization | 2.78 GiB | 3.30 GiB |
| Q4_K_M | — | 4.64 GiB |
| Licence | gemma | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | gemma-4-12B-coder-fable5-composer2.5-v1-bf16 | gemma-4-E4B-it | Ratio |
|---|---|---|---|
| 4,096 | 0.72 GiB | 0.12 GiB | 5.84× |
| 8,192 | 0.97 GiB | 0.18 GiB | 5.45× |
| 16,384 | 1.47 GiB | 0.29 GiB | 5.12× |
| 32,768 | 2.47 GiB | 0.51 GiB | 4.88× |
| 65,536 | 4.47 GiB | 0.94 GiB | 4.74× |
| 131,072 | 8.47 GiB | 1.82 GiB | 4.66× |