Model comparison
Mellum2-12B-A2.5B-Instruct vs gemma-4-E4B-it
At Q4_K_M, gemma-4-E4B-it is the smaller download — 4,977,171,584 bytes against 8,071,293,600.
From the file· summed bytes, KV per layer
Side by side
| Mellum2-12B-A2.5B-Instruct | gemma-4-E4B-it | |
|---|---|---|
| Parameters | 12.1B | 8.0B |
| Architecture | mellum | gemma4 |
| Layers | 28 | 42 |
| Native context | 131,072 | 131,072 |
| Mixture of experts | yes, 64 experts | no |
| Quantizations published | 2 | 36 |
| Smallest quantization | 7.52 GiB | 3.30 GiB |
| Q4_K_M | 7.52 GiB | 4.64 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Mellum2-12B-A2.5B-Instruct | gemma-4-E4B-it | Ratio |
|---|---|---|---|
| 4,096 | 0.12 GiB | 0.12 GiB | 1.06× |
| 8,192 | 0.17 GiB | 0.18 GiB | 1.04× |
| 16,384 | 0.28 GiB | 0.29 GiB | 1.02× |
| 32,768 | 0.50 GiB | 0.51 GiB | 1.01× |
| 65,536 | 0.94 GiB | 0.94 GiB | 1.01× |
| 131,072 | 1.81 GiB | 1.82 GiB | 1.00× |