Model comparison
Ministral-3-8B-Reasoning-2512-heretic vs gemma-4-12B-it
At Q4_K_M, Ministral-3-8B-Reasoning-2512-heretic is the smaller download — 5,198,385,440 bytes against 7,121,861,440. At long context the gap widens: gemma-4-12B-it's KV cache at 32K is 1.7× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| Ministral-3-8B-Reasoning-2512-heretic | gemma-4-12B-it | |
|---|---|---|
| Parameters | 8.9B | 12.0B |
| Architecture | mistral3 | gemma4 |
| Layers | 34 | 48 |
| Native context | 262,144 | 262,144 |
| Mixture of experts | no | no |
| Quantizations published | 3 | 59 |
| Smallest quantization | 4.84 GiB | 3.92 GiB |
| Q4_K_M | 4.84 GiB | 6.63 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Ministral-3-8B-Reasoning-2512-heretic | gemma-4-12B-it | Ratio |
|---|---|---|---|
| 4,096 | 0.53 GiB | 0.72 GiB | 1.35× |
| 8,192 | 1.06 GiB | 0.97 GiB | 1.10× |
| 16,384 | 2.13 GiB | 1.47 GiB | 1.45× |
| 32,768 | 4.25 GiB | 2.47 GiB | 1.72× |
| 65,536 | 8.50 GiB | 4.47 GiB | 1.90× |
| 131,072 | 17.00 GiB | 8.47 GiB | 2.01× |