Model comparison
gemma-4-12B-it-qat-q4_0-unquantized vs North-Mini-Code-1.0
These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: North-Mini-Code-1.0's KV cache at 32K is 2.2× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| gemma-4-12B-it-qat-q4_0-unquantized | North-Mini-Code-1.0 | |
|---|---|---|
| Parameters | 12.0B | 30.5B |
| Architecture | gemma4 | cohere2moe |
| Layers | 48 | 49 |
| Native context | 262,144 | 500,000 |
| Mixture of experts | no | yes, 128 experts |
| Quantizations published | 2 | 41 |
| Smallest quantization | 6.50 GiB | 7.93 GiB |
| Q4_K_M | — | 17.46 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | gemma-4-12B-it-qat-q4_0-unquantized | North-Mini-Code-1.0 | Ratio |
|---|---|---|---|
| 4,096 | 0.72 GiB | 0.38 GiB | 1.88× |
| 8,192 | 0.97 GiB | 0.52 GiB | 1.86× |
| 16,384 | 1.47 GiB | 0.72 GiB | 2.03× |
| 32,768 | 2.47 GiB | 1.13 GiB | 2.19× |
| 65,536 | 4.47 GiB | 1.94 GiB | 2.30× |
| 131,072 | 8.47 GiB | 3.57 GiB | 2.37× |