Model comparison
Ministral-3-8B-Reasoning-2512-heretic vs gemma-4-26B-A4B-it
At Q4_K_M, Ministral-3-8B-Reasoning-2512-heretic is the smaller download — 5,198,385,440 bytes against 16,796,017,248. At long context the gap widens: gemma-4-26B-A4B-it's KV cache at 32K is 2.8× 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-4-26B-A4B-it | |
|---|---|---|
| Parameters | 8.9B | 26.5B |
| Architecture | mistral3 | gemma4 |
| Layers | 34 | 30 |
| Native context | 262,144 | 262,144 |
| Mixture of experts | no | yes, 128 experts |
| Quantizations published | 3 | 46 |
| Smallest quantization | 4.84 GiB | 8.99 GiB |
| Q4_K_M | 4.84 GiB | 15.64 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-4-26B-A4B-it | Ratio |
|---|---|---|---|
| 4,096 | 0.53 GiB | 0.45 GiB | 1.18× |
| 8,192 | 1.06 GiB | 0.61 GiB | 1.75× |
| 16,384 | 2.13 GiB | 0.92 GiB | 2.31× |
| 32,768 | 4.25 GiB | 1.54 GiB | 2.75× |
| 65,536 | 8.50 GiB | 2.79 GiB | 3.04× |
| 131,072 | 17.00 GiB | 5.29 GiB | 3.21× |