Model comparison
Nethena-20B vs gemma-4-E4B-it
At Q4_K_M, gemma-4-E4B-it is the smaller download — 4,977,171,584 bytes against 12,042,208,800. At long context the gap widens: gemma-4-E4B-it's KV cache at 32K is 76.6× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| Nethena-20B | gemma-4-E4B-it | |
|---|---|---|
| Parameters | 20.0B | 8.0B |
| Architecture | llama | gemma4 |
| Layers | 62 | 42 |
| Native context | 4,096 | 131,072 |
| Mixture of experts | no | no |
| Quantizations published | 12 | 36 |
| Smallest quantization | 7.74 GiB | 3.30 GiB |
| Q4_K_M | 11.22 GiB | 4.64 GiB |
| Licence | cc-by-nc-4.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Nethena-20B | gemma-4-E4B-it | Ratio |
|---|---|---|---|
| 4,096 | 4.84 GiB | 0.12 GiB | 39.37× |
| 8,192 | 9.69 GiB | 0.18 GiB | 54.51× |
| 16,384 | 19.38 GiB | 0.29 GiB | 67.48× |
| 32,768 | 38.75 GiB | 0.51 GiB | 76.60× |
| 65,536 | 77.50 GiB | 0.94 GiB | 82.15× |
| 131,072 | 155.00 GiB | 1.82 GiB | 85.24× |