Model comparison
gemma-4-26B-A4B-it vs Ministral-3-8B-Reasoning-2512-heretic
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
| gemma-4-26B-A4B-it | Ministral-3-8B-Reasoning-2512-heretic | |
|---|---|---|
| Parameters | 26.5B | 8.9B |
| Architecture | gemma4 | mistral3 |
| Layers | 30 | 34 |
| Native context | 262,144 | 262,144 |
| Mixture of experts | yes, 128 experts | no |
| Quantizations published | 46 | 3 |
| Smallest quantization | 8.99 GiB | 4.84 GiB |
| Q4_K_M | 15.64 GiB | 4.84 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | gemma-4-26B-A4B-it | Ministral-3-8B-Reasoning-2512-heretic | Ratio |
|---|---|---|---|
| 4,096 | 0.45 GiB | 0.53 GiB | 1.18× |
| 8,192 | 0.61 GiB | 1.06 GiB | 1.75× |
| 16,384 | 0.92 GiB | 2.13 GiB | 2.31× |
| 32,768 | 1.54 GiB | 4.25 GiB | 2.75× |
| 65,536 | 2.79 GiB | 8.50 GiB | 3.04× |
| 131,072 | 5.29 GiB | 17.00 GiB | 3.21× |