Model comparison
Qwen3.8-27B vs Mixtral-8x22B-v0.1
At Q4_K_M, Qwen3.8-27B is the smaller download — 16,810,714,336 bytes against 85,586,760,928. At long context the gap widens: Qwen3.8-27B's KV cache at 32K is 3.5× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| Qwen3.8-27B | Mixtral-8x22B-v0.1 | |
|---|---|---|
| Parameters | 27.8B | 141B |
| Architecture | qwen35 | llama |
| Layers | 64 | 56 |
| Native context | 262,144 | 65,536 |
| Mixture of experts | no | yes, 8 experts |
| Quantizations published | 22 | 28 |
| Smallest quantization | 8.39 GiB | 16.99 GiB |
| Q4_K_M | 15.66 GiB | 79.71 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Qwen3.8-27B | Mixtral-8x22B-v0.1 | Ratio |
|---|---|---|---|
| 4,096 | 0.25 GiB | 0.88 GiB | 3.50× |
| 8,192 | 0.50 GiB | 1.75 GiB | 3.50× |
| 16,384 | 1.00 GiB | 3.50 GiB | 3.50× |
| 32,768 | 2.00 GiB | 7.00 GiB | 3.50× |
| 65,536 | 4.00 GiB | 14.00 GiB | 3.50× |
| 131,072 | 8.00 GiB | 28.00 GiB | 3.50× |