Model comparison
ZAYA1-8B-Coder vs gemma-4-E4B-it
At Q4_K_M, gemma-4-E4B-it is the smaller download — 4,977,171,584 bytes against 5,567,581,024. 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
| ZAYA1-8B-Coder | gemma-4-E4B-it | |
|---|---|---|
| Parameters | 8.8B | 8.0B |
| Architecture | zaya | gemma4 |
| Layers | 80 | 42 |
| Native context | 131,072 | 131,072 |
| Mixture of experts | yes, 16 experts | no |
| Quantizations published | 3 | 36 |
| Smallest quantization | 5.19 GiB | 3.30 GiB |
| Q4_K_M | 5.19 GiB | 4.64 GiB |
| Licence | — | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | ZAYA1-8B-Coder | gemma-4-E4B-it | Ratio |
|---|---|---|---|
| 4,096 | 0.31 GiB | 0.12 GiB | 2.54× |
| 8,192 | 0.63 GiB | 0.18 GiB | 3.52× |
| 16,384 | 1.25 GiB | 0.29 GiB | 4.35× |
| 32,768 | 2.50 GiB | 0.51 GiB | 4.94× |
| 65,536 | 5.00 GiB | 0.94 GiB | 5.30× |
| 131,072 | 10.00 GiB | 1.82 GiB | 5.50× |