Model comparison
Qwen3.6-35B-A3B 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 21,166,757,728. At long context the gap widens: Qwen3.6-35B-A3B's KV cache at 32K is 6.8× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| Qwen3.6-35B-A3B | Ministral-3-8B-Reasoning-2512-heretic | |
|---|---|---|
| Parameters | 36.0B | 8.9B |
| Architecture | qwen35moe | mistral3 |
| Layers | 40 | 34 |
| Native context | 262,144 | 262,144 |
| Mixture of experts | yes, 256 experts | no |
| Quantizations published | 61 | 3 |
| Smallest quantization | 8.77 GiB | 4.84 GiB |
| Q4_K_M | 19.71 GiB | 4.84 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Qwen3.6-35B-A3B | Ministral-3-8B-Reasoning-2512-heretic | Ratio |
|---|---|---|---|
| 4,096 | 0.08 GiB | 0.53 GiB | 6.80× |
| 8,192 | 0.16 GiB | 1.06 GiB | 6.80× |
| 16,384 | 0.31 GiB | 2.13 GiB | 6.80× |
| 32,768 | 0.63 GiB | 4.25 GiB | 6.80× |
| 65,536 | 1.25 GiB | 8.50 GiB | 6.80× |
| 131,072 | 2.50 GiB | 17.00 GiB | 6.80× |