Model comparison
Qwen3.8-27B vs North-Mini-Code-1.0
At Q4_K_M, Qwen3.8-27B is the smaller download — 16,810,714,336 bytes against 18,744,024,640. At long context the gap widens: North-Mini-Code-1.0's KV cache at 32K is 1.8× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| Qwen3.8-27B | North-Mini-Code-1.0 | |
|---|---|---|
| Parameters | 27.8B | 30.5B |
| Architecture | qwen35 | cohere2moe |
| Layers | 64 | 49 |
| Native context | 262,144 | 500,000 |
| Mixture of experts | no | yes, 128 experts |
| Quantizations published | 22 | 41 |
| Smallest quantization | 8.39 GiB | 7.93 GiB |
| Q4_K_M | 15.66 GiB | 17.46 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Qwen3.8-27B | North-Mini-Code-1.0 | Ratio |
|---|---|---|---|
| 4,096 | 0.25 GiB | 0.38 GiB | 1.53× |
| 8,192 | 0.50 GiB | 0.52 GiB | 1.04× |
| 16,384 | 1.00 GiB | 0.72 GiB | 1.38× |
| 32,768 | 2.00 GiB | 1.13 GiB | 1.77× |
| 65,536 | 4.00 GiB | 1.94 GiB | 2.06× |
| 131,072 | 8.00 GiB | 3.57 GiB | 2.24× |